interactivity and enaction in human cognition ... · cognitive s c ien c e con c ept s in e na c...

Cogn

itiv

e SC

ienC

e Co

nCep

tS in

ena

Ctiv

iSm

234

interactivity and enaction in Human Cognition

ConstruCtivist Foundations vol. 11, n°2

enactivism

interactivity and enaction in Human CognitionMatthew Isaac Harvey • University of Southern Denmark • harvey/at/sdu.dkRasmus Gahrn-Andersen • University of Southern Denmark • rga/at/sdu.dkSune Vork Steffensen • University of Southern Denmark • s.v.steffensen/at/sdu.dk

> Context • Distributed language and interactivity are central members of a set of concepts that are rapidly develop-ing into rigorous, exciting additions to 4E cognitive science. Because they share certain assumptions and methodologi-cal commitments with enactivism, the two have sometimes been confused; additionally, while enactivism is a well-developed paradigm, interactivity has relied more on methodological development and on a set of focal examples. > problem • The goal of this article is to clarify the core conceptual commitments of both interactivity-based and enactive approaches to cognitive science by contrasting the two and highlighting their differences in assumptions, focus, and explanatory strategies. > method • We begin with the shared commitments of interactivity and enactivism – e.g., antirepresentationalism, naturalism, interdisciplinarity, the importance of biology, etc. We then give an overview of several important varieties of enactivism, including sensorimotor and anti-representationalist enactivism, and then walk through the history of the “core” varieties, taking care to contrast Maturana’s approach with that of Varela and the current researchers following in Varela’s footsteps. We then describe the differences between this latter group and interactivity-based approaches to cognitive science. > Results • We argue that enactivism’s core concepts are explanatorily inadequate in two ways. First, they mis-portray the organization of many living systems, which are not operationally closed. Second, they fail to realize that most epistemic activity (i.e., “sense-making”) depends on engage-ment with non-local resources. Both problems can be dealt with by adopting an interactivity-based perspective, in which agency and cognition are fundamentally distributed and involve integration of non-local resources into the lo-cal coupling of organism and environment. > implications • The article’s primary goal is theoretical clarification and exposition; its primary implication is that enactive concepts need to be modified or extended in some way in order to explain fully many aspects of cognition and directed biological activity. Or, read another way, the article’s primary im-plication is that interactivity already provides a rich set of concepts for doing just that, which, while closely allied with enactivism in several ways, are not enactivist concepts. > Constructivist content • The article consists entirely of a comparison between two constructivist fields of theory. > Key Words • Interactivity, enactivism, distributed language, radical embodied cognitive science, ecological psychology, autonomy.

introduction

« 1 » This target article is an effort to come to grips with the relation between in-teractivity and enaction. The first of these has not been systematically formulated as a distinctive paradigm, but is nonetheless emerging as a vital set of concepts underly-ing research on cognition and social inter-action. This body of research, which we will provisionally refer to as “interactivity-based approaches,” explores ecological, distribut-ed, and bodily aspects of human action and social inter-action (Cowley 2011; Cowley & vallée-tourangeau 2013; steffensen 2015). The fundamental commitments of the sec-ond – an increasingly influential paradigm for research in cognitive science – are to the inseparability of explanations for cognition

from explanations for life, and to the self-producing and self-sustaining organization of living systems as a resource for explana-tions of both kinds (stewart, Gapenne & di Paolo 2010; Thompson 2007). The two clusters of concepts are fairly close relatives because they share many theoretical pre-decessors. The research programs in which they are involved are also natural allies, as both are actively pushing the explanatory boundaries of non-representational cogni-tive science, and both take for granted the fundamental importance of embodiment, biological dynamics, and ecological embed-dedness in accounting for cognition and agency.

« 2 » although the development of en-active theory has been a source of insight and inspiration for our interactivity-based

approach, it is not a variety or descendant of enactivism. our interest in social agency, interaction, and cultural aspects of human life are a poor match for some of the core te-nets and concepts in enactivist thought, and we are not fully satisfied with the resources enactivism provides to answer questions in these domains. This dissatisfaction comes primarily from our inability to find, in ex-isting enactivist accounts of social cognition (e.g., Cuffari, di Paolo & de Jaegher 2014; de Jaegher & di Paolo 2007; de Jaegher, di Paolo & Gallagher 2010; de Jaegher & Froese 2009; Froese & di Paolo 2009, 2010, 2011) any account of the ecological transfor-mation that languaging and human sociality bring about, which also conditions individu-al cognitive activity. This lack has its roots in a more fundamental mismatch between en-

http://constructivist.info/11/2


235

interactivity and enaction in Human Cognition matthew isaac Harvey et al.

Enactivism

http://constructivist.info/11/2/234.harvey

activist and interactivity-based assumptions about the basic nature of relations between agents and their environments. in general, where enactivism describes agent-environ-ment relations in terms of in-the-moment coupling, we assume these relations play out on multiple heterogeneous timescales, such that in many cases – perhaps most – agency is fundamentally distributed.

« 3 » This difference is most easily brought out in the following way. When one begins with human action as an object of study, it is immediately apparent that it can-not be exhaustively explained in terms of the coupling of bodily movement to surfaces of the environment. This is usually because the action being studied in some way di-rectly depends on customs, objects, expe-riences, and events that are outside its im-mediate spatiotemporal scope. For instance, re-conceptualization can lead someone to perceive previously unnoticed affordances of an otherwise unchanged material set-ting, or the discovery of small, unexpected sensorimotor contingencies while interact-ing with technology can refocus their atten-tion, their efforts, or their thinking about the situation. “interactivity” arose as a way of talking about the potential for this type of person-environment engagement pos-sessed by particular artifacts, or spaces, or problems, and has in general been used to refer to a research interest rather than a spe-cific theoretical construct. its original and most frequent use is in the study of human-computer interaction, where it picks out an interest in the small-scale contingencies of action–perception cycles in coupling to technological interfaces (Johnson, Bruner ii & Kumar 2006; McMillan 2006). More recently, it has been borrowed by scholars interested in distributed cognition to name a related, loosely-defined, class of phenom-ena that includes the small-scale dynamics of engagement with artifacts and other peo-ple in problem solving, professional settings such as offices and emergency rooms, and more bodily practices such as choreography and jewelry-making (Baber, Parekh & Cen-giz 2014; Cowley & nash 2013; Cowley & vallée-tourangeau 2013; Kirsh 1997, 2015; Pedersen 2012).

« 4 » development of interactivity as a concept has been slow to follow its use as a topical designator, but several recent papers

(steffensen 2015; steffensen & Pedersen 2014; steffensen, vallée-tourangeau & val-lée-tourangeau 2015) have made headway in this regard, linking interactivity to multi-scalar dynamics of the cultural, historical, and ontogenetic contexts in which human action is embedded. By way of contrast with enactivism, approaches based on this notion of interactivity begin with the ecological embedding and social normativity of hu-man action, and then seek to understand its biological nature, rather than looking for purely biological explanatory principles.

« 5 » The rest of the article takes the fol-lowing course. in the next section, we will lay out our understanding of enactivism as a research paradigm, noting points of agree-ment and detailing the enactivist position on issues important to our comparison, which have primarily to do with varela-inspired concepts such as autonomy, adap-tivity, and sense-making. The subsequent section presents our concerns with that po-sition, which are

� that it mis-characterizes the organiza-tion of many living systems by inad-equately accounting for the constitutive role of environmental structures and socially normative patterns of behavior, and

� that it incorrectly ascribes sense-making to intrinsic teleology on the time-scale of an organism’s biological self-mainte-nance, which we take to be far too nar-row a characterization of the relevant sources of constraint and normativity.

in the section “interactivity,” we build up a clearer notion of interactivity as sense-satu-rated coordination that enables and contrib-utes to human action, and then explore how this notion can be made both richer and more precise. We conclude with a summary of the points of agreement and divergence between our position and post-varela au-topoietic enactivism.

varieties of enactivism

« 6 » The terms “enaction” and “enac-tive” were first used in something like their current sense in Francisco varela, Evan Thompson, and Eleanor rosch’s (1991) The Embodied Mind. That book was both influ-ential and to a large extent successful in its

effort to place first-person experience and systems-theoretic biology at the cutting edge of “the new sciences of mind,” by which bold phrase the authors meant the inchoate collection of ideas and methods that were opposed to the cognitivist tradition in cog-nitive science. These are now widely referred to by the catch-all label “4E cognition,” where the four Es indicate interest in em-bodied, embedded, extended, and enacted aspects of the function of cognitive agents (Menary 2010; rowlands 2010). as we see it, “enactivism” is a heterogeneous subset of contemporary 4E approaches to the study of mind and agency, distinguished either by their acceptance of the notions of autopoi-esis and autonomy (e.g., Barandiaran & Eg-bert 2014; Barandiaran, di Paolo & rohde 2009; di Paolo 2009b; Froese & Gallagher 2012; Froese & Ziemke 2009; Froese, virgo & ikegami 2014), due to varela’s work with Humberto Maturana (Maturana & varela 1980, 1987; varela 1979), or by their concern with phenomenology and lived experience (e.g., Colombetti 2014; degenaar & Myin 2014; Gallagher & Hutto 2008; noë 2009; ratcliffe 2012), or both.1

« 7 » Many of these approaches are his-torically non-central varieties of enactiv-ism in that they are (or are in part) direct intellectual descendants of varela, Thomp-son, and rosch, but are not focused on developing their central ideas. Being in a similar theoretical situation ourselves, we are sympathetic to such approaches; two in particular – often called “sensorimotor” and “radical” enactivism, respectively – are help-ful in spelling out our own position. after briefly discussing these, we will turn toward the paradigm’s “primary literature,” com-posed of the original work of Maturana and varela (identified as such by e.g., di Paolo 2005, Froese & stewart 2012). our overview will follow a roughly chronological course, beginning with Maturana’s biology of cogni-tion and the original notion of autopoiesis, which we understand as a non-explanatory abstract description of the organizational characteristics of living systems. varela’s

1 | We will use the phrase “lived experience” to indicate the phenomenal, experiential char-acter of organisms’ engagements with the world, from their own “first-person” point of view.



Cogn

itiv

e SC

ienC

e Co

nCep

tS in

ena

Ctiv

iSm

236


Enactivism



Enactivism

work diverges from Maturana’s in at least two ways:2

a it moves towards formally rigorous systems-theoretic accounts of autonomy and so also towards a conception of au-topoiesis as a specific sub-category of autonomous systems; and

b it introduces the ideas of intrinsic tele-ology and normativity, and enaction as sense-making (from which comes the idea of “life-mind continuity”).

These changes formed the basis for the logi-cal extension of autopoiesis that we refer to here as “post-varela autopoietic enactivism,” and which is exemplified by the work of Ezequiel di Paolo, tom Froese, Hanne de Jaegher, John stewart, Xabier Barandiaran, and others. This section is intended expo-sitionally; we will argue in subsequent sec-tions that both aspects of the turn towards post-varela enactivism are problematic, and also that they introduced claims that are in-compatible with interactivity in ways that Maturana’s original work is not.

Sensorimotor enactivism« 8 » one of the best-known strands of

enactive research concerns the inseparabili-ty of action and perception. The core insight of “sensorimotor enactivism” (o’regan & noë 2001a, 2001b) is that for organisms with nervous systems, motion is always di-rectly regulated by sensation, which is itself structured and given its character by the dynamics of bodily movement in relation to the organism’s environment (for overviews, see Hurley 2002; noë 2004; o’regan 2011). Perception is active and virtual, by which it is meant that it depends not only on cur-rent sensation but also on the way sensation, and current movement, is systematically related to upcoming and potential sensory events, on the basis of the structure of the environment and the organism’s neuromus-cular expertise (Mossio & taraborelli 2008; noë 2012). This expertise comprises an-ticipatory sensitivity to the contingencies of sensory events in relation to potential

2 | it is arguable whether the issues we iden-tify here reflect an actual difference in opinion between the two academics. at the very least, however, there are two distinct conceptual posi-tions, which seem to correspond roughly to their separate views.

or ongoing bodily movement, and is devel-oped by means of individually and socially regulated experiences during ontogeny (for discussion, see Buhrmann & di Paolo 2014: 21–31; Buhrmann, di Paolo & Barandiaran 2013; Hutto 2005; Loughlin 2014).

« 9 » sensorimotor theory is enactivist in at least three ways. First, it contradicts the cognitivist model of perception as the input of information, which, after being suitably processed, might be used to choose a suitable course of action-output. in the sensorimotor enactivist view, action and perception are inextricable, and perception constitutively involves both sensorimotor skill (which is therefore not a separate type of non-perceptual information processing) and elements of the environment that partly specify sensorimotor contingencies (see e.g., Barandiaran 2012; Hurley 2001; Gan-gopadhyay & Kiverstein 2008; Prinz 2006). second, the approach is centrally concerned with first-person lived experience, which it understands in terms of the active use or ap-plication of bodily sensorimotor expertise to engage with the world; this makes it closely allied with the phenomenological aspect of varelian enactivism (Myin & o’regan 2002; o’regan, Myin & noë 2005). and third, given these two similarities, sensorimotor theory can be understood as a partial inter-pretation, in concrete biological terms, of the core idea that organisms “enact,” or ac-tively construct, the world from their indi-vidual perspective (noë 2012: 58f; o’regan & noë 2001a: 945; Thompson 2005, 2007: 254–265).

« 10 » our understanding of sensorim-otor enactivism is that it is a description of how organisms with nervous systems en-gage perceptually with their worlds. We take it to be a non-exhaustive and very general but nonetheless scientifically robust account of the mechanisms involved, one that we see no reason to reject and every reason to support. in fact, the notions of interactivity, distributed cognition, and so on are very dif-ficult to make sense of without some version of a sensorimotor account of perception and action (Cowley 2009; Cowley & vallée-tourangeau 2013; rączaszek-Leonardi & Cowley 2012; Thibault 2011).

enactive anti-representationalism« 11 » We are similarly in almost com-

plete agreement with “radical enactivism,” as described and defended by daniel Hutto and Erik Myin (2013; see also Hutto, Kirch-hoff & Myin 2014; Hutto 2013; Hutto 2015). Their “enactivism” is self-professed, and consists entirely of the claims that:a agency and cognition are fundamentally

a matter of dynamical bodily engage-ment with the world; and

b the notions of content and representa-tion are not naturalistic and so cannot be used in naturalistic explanations.

The first of these is a straightforward com-mitment to the embodiment of cognition rather than a strictly enactivist thesis, and so the authors write that they “remain neutral toward other, more extravagant claims as-sociated with the original version of enac-tivism” (Hutto & Myin 2013: 5). The second claim amounts to a consistent, technically sophisticated philosophical anti-representa-tionalism, whose core is the “Hard Problem of Content.” it can be informally summa-rized as follows:

� any valid explanation of cognition must be naturalistic.

� The idea of “representation” minimally requires positing a state or structure that bears “content.”

� Therefore representational theories of cognition are predicated on the claim that content is a naturalistic property or substance.

� all serious attempts to account for con-tent in naturalistic terms have involved treating it as a kind of information, but

� information-as-covariance (which is the only naturalistic kind of information3) lacks the minimal properties necessary for something to count as content.

� Conclusion: representationalism that relies on information-as-content is not naturalistic, and therefore cannot offer valid explanations for cognition.

3 | This is probably the most contentious part of the argument, as many philosophers follow Fred dretske (1988) and ruth Millikan (2004) in holding that there are richer, equally natural-istic theories of information. For discussion, see Miłkowski (2015a, 2015b) and Myin & Hutto (2015).




Enactivism

237


Enactivism


« 12 » interactivity-based approaches are premised on a similar anti-representa-tionalist stance, albeit one that is motivated as much by concerns about explanatory util-ity as by concerns about philosophical sta-tus. Because of this difference in theoretical motivation, our anti-representationalism is more thoroughgoing than that of “radical enactivism.” Hutto and Myin hold, as we do not, that any cognitive activity that is both phylogenetically and ontogenetically subse-quent to the development of “natural human language” is likely to be representational and content-involving (e.g., Hutto & Myin 2013: 41, 151–153; see Harvey 2015 for in-depth discussion, and cf. Hutto & satne 2015; Kiverstein & rietveld 2015). We agree with them that the directedness of activity inher-ent in biological self-maintenance (di Paolo 2005; Thompson 2007: 140–148; Weber & varela 2002) is not of the rule-following type usually described as “normativity” and that is often taken to characterize language (Hutto & Myin 2013: ch. 2; cf. de Jesus 2015; Heras-Escribano, noble & de Pinedo 2014). and further, we agree that conventions, rules, and so forth are uniquely social and have to do with inter-agent coordination. But it does not seem to us that this justifies calling them representational or contentful. on the contrary: one of our core assump-tions is that “representation” and associated concepts are exactly as explanatorily vacu-ous when invoked to account for socially situated activity as they are when invoked to account for the behavior of simple organ-isms.4

maturana« 13 » Here we turn to the historical

roots of enactivism. in the 1970s and 80s, Humberto Maturana articulated a series of concepts that he regarded as useful in un-derstanding the relation of biology to hu-man knowing and engagement with the world. very roughly, these include an episte-mology and implicit ontology (the observer, structure determinism), a model for scien-tific explanation (in terms of distinctions,

4 | We agree that certain phenomena – imag-ination, dreaming, and memory, for instance – are self-evidently representational in some (loose and descriptive) sense, but this does not make the con-cept of representation an explanatorily useful one.

unities, and generative mechanisms for phenomena), an abstract characterization of the organization of biological systems at the molecular level (autopoiesis), and an equally abstract conception of cognition in terms of organisms’ experience of being coupled to the world. some of these lose their intended meaning if considered apart from the oth-ers; in particular, autopoiesis must be un-derstood in terms of Maturana’s model for scientific explanations. The concept of auto-poiesis

� specifies the organization of living sys-tems, and

� grounds and justifies Maturana’s ontolo-gy and closely connected epistemology.« 14 » it is necessary to begin by distin-

guishing “organization” from “structure,” on the one hand, and from “operation,” on the other. organization is an abstract specifica-tion of the set of relations that obtain among the system’s repeating component processes so long as the system persists. a system’s organization produces the properties that define it as a member of a particular type or class of system for an observer, and also “defines it as a composite unity.” structure is the set of processes or components that in fact realize a given system’s organization at a particular point in time (Maturana 1978: 32f; Maturana & varela 1980, 1987: 42–52;5 varela 1979: 8–11). The way Maturana uses the terms, both organization and structure are defined as abstractly immediate, which is to say, it is assumed that the relations com-prising a given organization or structure are dynamic ones, but the scales, rates, and so forth on which they play out are not speci-fied. For this reason, it is important that we separate a system’s (static) organization from its operations, that is, the actual realiza-tion of its component processes as they play out over time, and their causal effects on one another and on system-external processes.6

5 | several early works co-authored by Mat-urana and varela (e.g., 1980, 1987; varela, Mat-urana & uribe 1974) take a distinctly “Matura-nian” position. varela moves away from Maturana in his individual work.

6 | This distinction is not drawn by Maturana and varela, but it is necessary to make sense of the differences between them. it is coherently articu-lated both by david rudrauf, antione Lutz, diego Cosmelli, Jean-Philippe Lechaux, and Michel le

« 15 » autopoiesis is an organization, and autopoietic(ally organized) systems are networks of processes of molecular produc-tion, transformation, and disintegration, which processes, as they take place, con-tinually re-produce the components of the network such that it both persists in time and defines itself as a coherent spatial unity (Maturana & varela 1980: 78f). The para-digm case is the animal cell, whose internal (metabolic) processes produce the mole-cules needed to make up the cell membrane, which in turn admits only certain molecules into the cell’s interior, thereby enabling the internal production to continue (varela, Maturana & uribe 1974). autopoiesis is not an explanatory concept; it is instead an “ab-straction” of molecular dynamics as a con-ceptual set of relations (Maturana 2002m 2011).7

« 16 » The ontological and epistemo-logical nature of the concept comes from the additional notion of “autonomy,” which is introduced to thematize the fact that mo-lecular events that affect a cell (without de-stroying it) will have their effects according to the cell’s self-production. Whatever the effects are – a change in cytoplasmic salin-ity to match the salinity of the extracellular milieu, say – they will be determined by the way the change affects the cell’s auto-poietic organization (Maturana 1970: 17f, 1978: 36f). “autonomy” gives a name to this partial organizational independence; its conceptual purpose is only to highlight an aspect of autopoiesis (Maturana & Poerk-sen 2004: 68–81).8 When Maturana says

van Quyen (2003: 34), and by Thompson (2007: 448n4). virgo, Egbert, and Froese (2011: 243) similarly distinguish the processes comprising a system from the system dynamics they underlie.

7 | Maturana’s understanding of scientific explanation is that it involves “intended repro-duction or reformulation of a system or phe-nomenon,” and so to propose an explanation is to “propose conceptual or concrete systems that can be deemed to be intentionally isomorphic to (models of) the systems that generate the ob-served phenomena” (Maturana 1978: 29). The no-tion of autopoiesis is a conceptual system of this intentionally reproduced kind.

8 | Thus we read that “a system is autono-mous if it can specify its own laws, what is proper to it” (Maturana & varela 1987: 48). also in the



Cogn

itiv

e SC

ienC

e Co

nCep

tS in

ena

Ctiv

iSm

238


Enactivism



Enactivism

(as he does frequently) that autonomy is a more general notion than autopoiesis, he means that autopoiesis is a description of the reason why (and the way in which) liv-ing systems are autonomous in this general sense (Maturana 2011). so the autonomy of living systems does not amount to their static molecular self-organization regulat-ing all constitutive operational dynamics of the system. it amounts, instead, to the claim that no external constraint can affect its ac-tivity in a way that is somehow irrespective of the interdependence among its compo-nent molecular processes.

« 17 » This point about autonomy brings up a more general aspect of Mat-urana’s work, which is that because it is strictly abstract, strictly synchronic, and bound up with non-systems-theoretic epis-temological and ontological assertions, it makes no particular claims about the op-erational dynamics of living systems. For instance, Maturana writes of an organism’s surroundings simply as its “medium,” not bothering to differentiate it or speak of its structure or role; what matters to him is only that that structure will always affect the agent in terms of the agent’s organiza-tion. and so it seems to us that examples of circular molecular organization might, without contradiction, be realized opera-tionally by non-circular, constitutively open and environment-involving, massively dis-tributed and multi-scalar processes. For in-stance, both eating and breathing are ways of maintaining organizational closure on the molecular level, but they do so irregu-larly, over far longer timescales than those on which metabolic processes play out. When we refer to Maturana in the rest of the essay, we will have in mind his position construed in this way.

Poerkson interviews, human autonomy is not equated with operational self-sufficiency but rather with moral responsibility (Maturana & Poerksen 2004: 74f). and, in his initial paper on the biology of cognition, the word “autonomy” appears only once, in the post-scriptum, where Maturana writes, “man is a deterministic and rel-ativistic self-referring autonomous system whose life acquires its peculiar dimension through self-consciousness” (Maturana 1970: 26).

varela« 18 » as di Paolo (2009a: 11) tells us,

“enaction, varela-style, starts with the ques-tion of the autonomy of the cognizer.” al-though they worked together very closely, varela’s approach to autopoiesis and au-tonomy seems never to have been perfectly aligned with Maturana’s. as early as 1979, varela was focusing on autonomy as an operationally defined systems-theoretic concept. That is, he understood it as a claim about the dynamics of a system as it real-ized itself over time, and also as an aspect of the systems-theoretic composition of organisms, rather than as a way of talking about a basic, non-generalizable condition of all living.

« 19 » There are several elements of this change (see varela 1979: 52–57). First, varela defines autonomy as the ability of a system to delineate itself in some domain by means of its “organization and opera-tion,” in such a way that an observer can identify it through coupling to it in that same domain. For instance, we can identify cells through being coupled to them in the molecular domain. second, varela defines circular self-realization in terms of (what we would call, given the terminology we are using here) operational closure, rather than closure of static organizational relations. to quote him at length:

“ We shall say that autonomous systems are organizationally [in our terms, ‘operationally’] closed. That is, their organization is characterized by processes such that (1) the processes are related as a network, so that they recursively depend on each other in the generation and realization of the processes themselves, and (2) they constitute the system as a unity recognizable in the space (do-main) in which the processes exist…once this circularity arises, the processes constitute a self-computing organization, which attains coherence through its own operation, and not through the intervention of contingencies from the environ-ment.” (varela 1979: 55, emphasis added)

« 20 » This reflects a general shift to thinking of autonomy as a technical con-cept for the analysis of systems, a shift that is realized in various ways. For instance, varela (1979: ch. 11–13) expends signifi-cant effort in developing a formalization of autonomy in systems-theoretic terms,

and for the same reason, he is led to treat “autopoiesis” as a formal specification of a subset of autonomous systems, and au-tonomy as the general case of autopoiesis. He suggests that examples of other types of autonomous systems might include so-cial units such as families, companies, and ecosystems, provided that they display operational closure in some appropriate non-physical domain (ibid: 53).9 all of this supports our conclusion that, in the above passage, varela is suggesting that autono-mous systems are not only organizationally closed in the static, abstract, and non-spe-cific sense proposed by Maturana, but also closed in their actual dynamic processes of self-realization (i.e., their operation). The interpretive move we made with Maturana – to construe his organizational account as compatible with a wide range of distributed and heterogeneous operational dynamics – is thus not open to him.

« 21 » varela’s thought also moves away from Maturana on the relation of cognition and experience to autopoiesis. in his later work, he developed a view of autonomous systems as immanently (i.e., intrinsically) purposive, in the sense that if a network of processes is self-realizing, the operations of those processes must – as a matter of neces-sity – be such that they maintain the overall organization of the system. This idea comes from varela’s application of Kantian philoso-phy to his pre-existing views on autonomy and autopoiesis, a process summarized in andreas Weber and varela (2002). Thomp-son (2007: 146) explains that over time, varela “came to believe that this notion of immanent purposiveness is not simply de-scriptive but explanatory,” in that it helps to educe an additional aspect of the logic of autonomy. For a system to be autonomous is for it to create itself as a stable “identity” among fluctuations of matter and energy. to

9 | it should be noted that Maturana also wrote about “second-order” (multicellular) and “third-order” (multi-organism) autopoiesis, as does varela, but in those accounts (e.g., Maturana & varela 1987: 87–89; varela 1979: 50–52) it is not clear whether they are describing a further degree of abstraction away from the operational details of molecular self-organization, or whether they are describing something like the technical varelian notion of autonomy.




Enactivism

239


Enactivism


have an identity is to have established “a logi-cal and operational reference point” for one’s engagement with the world, which engage-ment is therefore (again, necessarily) con-ducted in accordance with the needs specific to one’s particular manner of self-creation (varela 1991, 1992, 1997). varela used the term “sense-making” for engaging with the world in this intrinsically intentional way.

« 22 » so in our view, varela’s work rep-resents two trends away from Maturana. These are:

� a new understanding of “autonomy” in technical, systems-theoretic terms, which leads to a construal of autopoiesis as a subset of autonomy; and

� a model of cognition as the way an or-ganism’s (or agent’s, or person’s) opera-tional closure constrains its interactions with its environment.

post-varela autopoietic enactivism« 23 » Both of these trends are carried

forward by the strand of thought that we call “post-varela autopoietic enactivism,” which is what we will refer to from here on when we use the word “enactivism.” For instance, ruiz-Mirazo and Moreno (2012) refer to varelian autonomy as “the core of life” and see it as the key to explaining evo-lution. Thompson (2007: 44) places varela’s account of autonomy at the heart of his ac-count of cognition. and for de Jaegher and di Paolo (2007), coupling between autono-mous agents is the basis for explanations of social interaction. The clearest development of these trends is a line of research that be-gins with di Paolo’s (2005) observation that when construed in formal, systems-theoret-ic terms, autonomy is actually insufficient to entail the logic of sense-making (see also Thompson 2004). in particular, a network of processes that displays bare self-main-tenance – call this “minimal autonomy” – does not, simply by virtue of that property, have the ability to continue to maintain itself when perturbed, or when its environmental conditions change.

« 24 » This observation leads di Paolo to propose a hierarchy of increasingly rich systems-theoretic concepts to account for increasingly complex ways that living systems can relate to their environments, which, it should be noted, all describe op-erational characteristics rather than organi-

zational ones (for a detailed presentation of these see di Paolo 2015; di Paolo, rohde & de Jaegher 2010; Froese & di Paolo 2009, 2011). “adaptivity” describes the minimal case of a system that is autonomous and, in addition, can respond to perturbations by modifying its operations so as to keep stable some homeostatic or homeodynamic vari-able. This selective responsiveness is mini-mal “sense-making.” “agency” – or minimal behavior – is adaptivity achieved by means of a system regulating some aspect of its coupling with the environment external to its operational bounds, which it must do ac-cording to a “value” or “norm” generated by its manner of self-production. “Mentality” requires that the coupling so regulated be sensorimotor, and “sociality” requires that the system’s intrinsic norms are underde-termined by metabolism (i.e., they concern maintenance of system identities other than bare molecular identity).

Difficulties with agency and sense-making« 25 » in this section, we articulate

several concerns about the varelian and post-varelian enactivist approaches, hav-ing to do with agency in general and with the concept of sense-making in particular. But first, let us be clear about what we do not take issue with, as many of the enactive approach’s central commitments are shared by our interactivity-based approach. in ad-dition to our agreement with sensorimotor explanations for perception and with thor-oughgoing anti-representationalism, we concur that the study of living systems can and should involve the effort to character-ize their organization as well as the dynam-ics of their biochemical operations. We are convinced – by enactivist research as well as by work in related fields (e.g., Baber, Parekh & Cengiz 2014; Bailly & Longo 2008; Lal-and, odling-smee & Feldman 2000; turvey 2007; van orden, Holden & turvey 2003) – that such characterizations comprise a useful, productive, and scientifically valid approach to the study of organisms, agency, and cognition. additionally, we take it that the lived experience of agents is crucial to any viable account of cognition, human or otherwise, and we wholeheartedly agree

with di Paolo (2005) that cognitive science must use interdisciplinary methods and treat its many objects of study as inherently interrelated. With all of this in mind, there are two specific issues on which our posi-tion diverges sharply from that of the enac-tivist paradigm.

the operational openness of living systems« 26 » our first concern is that most

organisms, perhaps all, achieve molecular self-realization in a way that constitutively involves body-environment interactions. That is, (a) often the activity of organisms is regulated by influences that are not inher-ent in either their synchronically defined organization, their immediate sensory ar-ray (i.e., that are not immediately present as impingements or perturbations on the system’s autonomy), or the emergent au-tonomous dynamics of ongoing organism-environment coupling. and (b) organisms rely on these regulatory influences to stay alive, and use them as basic, lineage-general ways of maintaining their molecular self-reproduction. now, if these situations are simple cases of an organism regulating an environmental feature such that the feature became an enabling condition for some of its body-internal processes, then they would simply involve a mis-match between the spa-tial and operational bounds of the organism, which is perfectly compatible with the enac-tivist framework and its model of organism-environment asymmetry (virgo, Egbert & Froese 2011). similarly, if they only involve a one-way, transient enabling relation between an organism-external and an organism-in-ternal process, there would be no reason to doubt the underlying stability of operational closure. But it is doubtful whether an opera-tional closure-based account can describe even straightforward phenomena such as nest-building and web-building, where agents’ basic methods of thermoregulation, orienting themselves in space, and internal-izing molecules required for metabolism are organized on timescales other than those of bodily motion and molecular production (sterelny 2010). For instance, if members of a species can only reach maturity by growing inside a cocoon, surely past acts of cocoon-building are constitutive elements of their maturational processes. our first contention,



Cogn

itiv

e SC

ienC

e Co

nCep

tS in

ena

Ctiv

iSm

240


Enactivism



Enactivism

then, is this: the logic of varelian autonomy does not work for situations other than monoscalar coupling between an organism and a non-living milieu of some kind.10

« 27 » The mistake here, as we under-stand it, is in making an unnoticed jump from autopoietic organizational closure to substrate-general operational closure. The first notion is a Maturana-inspired view along these lines: the molecular processes of synthesis and breakdown responsible for creating the body of any given organism comprise a circularly organized network through their relations of enablement, al-though the way they enable one another may be indirect (i.e., may involve various other mediating processes that are not processes of molecular production or immediately relevant thereto, and may play out on dif-ferent spatial and temporal scales). By con-trast, substrate-general operational closure (varelian “autonomy”) describes a situation where the dynamics of the processes con-stituting an organism – including processes other than molecular production – depend on and condition one another, and enable one another directly, that is, enable one an-other without depending on non-systemic processes. Whereas (our understanding of Maturana’s conception of) autopoiesis is an abstract conceptual reconstruction of a particular aspect of the way living systems function, and can be realized in a wide va-riety of ways, varelian autonomy is a char-acterization of the defining properties of the system’s actual operational dependen-cies over time. This is why di Paolo argues that it has to be augmented with additional requirements and details if it is to be accu-rately applied to complex organisms. He is correct insofar as autonomy is not an ac-curate characterization of the operational dependencies among most organisms’ com-ponent processes, but incorrect in that he assumes that autonomy always obtains, and that what is needed are details of what the organism does in addition to realizing its operational closure.

10 | This is a generalization and extension of the criticism in Cowley and Gahrn-andersen (2015), which is that autonomy is incompatible with the way that social animals rely on one an-other – and on history, and traditions, and envi-ronmental structures – to organize their actions.

« 28 » it should be noted that both or-ganizational closure and non-autonomous living systems have been recognized as the-oretical possibilities in the enactivist litera-ture. nathaniel virgo and colleagues define autonomy in terms of a-temporal closure of enabling conditions among repeating proc-esses, rather than in terms of synchronic closure of enabling conditions among proc-esses as they play out in real time. and so they write,

“ on this view the autopoietic system that consti-tutes the [animal] is not coextensive with the unity that we refer to as ‘[the animal],’ it is much bigger. This will be the case for most if not all organisms, since most organisms rely not only on sensory-motor loops that run through their environment but also on nutrients that are recycled externally to them.” (virgo, Egbert & Froese 2011: 246)

if we are careful, we will see that this is a species of organizational closure as we have defined that concept, although their ver-sion strongly implies that the processes in question are all on the same timescale and does not allow for constitutive openness of operations. This latter possibility is ac-knowledged, instead, by di Paolo (2009b: 63–66), for whom it arises as the idea of a system that is alive without strictly being autonomous. such a system would be one whose self-production (i.e., metabolism) is organized such that it could only persist in the presence of enabling conditions (e.g., the presence of nutrients) provided by processes of environmental interaction rather than by processes inside the organism’s operational bounds. Where this description applies, the system is literally constituted by environ-mental interaction. di Paolo uses a fictional animal to illustrate his point, but any animal capable of taxis would have done just as well: our claim is precisely that this obtains for almost all organisms, and that it should be the starting point for research rather than a mere logical possibility.

« 29 » it is helpful to consider an ex-ample in more detail. scott turner (2000: 41–51) describes “bioconvection,” in which convection cells emerge in the collective movement of large groups of microorgan-isms. He uses the example of Chlamydo-monas nivalis, which swim straight ahead whenever they are in water that lacks suffi-

cient oxygen for their continuing metabolic function, and whose bodies are asymmetri-cal such that in a still fluid, their massier back ends rotate down beneath them, with the effect that they swim upwards. in an open culture, where oxygen enters the water from the air and slowly diffuses downwards, Chlamydomonas cells will swim upwards en masse, concentrating near the surface. This situation is unstable, as it is effectively a fluid whose top layer is much denser than the layers beneath it, so areas with a particu-larly high concentration of cells will begin to sink. as this “anti-bubble” of microor-ganisms falls through the fluid, it creates a column of downward-flowing water, which has the effect of rotating the nearby Chlam-ydomonas cells towards the center of the col-umn (again, because their “back” ends have greater mass than their “fronts”), into which their swimming then carries them. This in-creases the density of the fluid in the middle of the column, causing it to continue to sink, and making the downward-flowing column of fluid into a fairly stable pattern. at the bottom of the culture, the cells disperse and start to swim upwards again, while at the surface, cells are “falling” into the column and drifting downwards. The net effect is that in a suitable environment, simply by swimming forward, a colony of Chlamydo-monas is able to provide all of its member cells with adequate access to oxygen-rich surface water, allowing their cellular (e.g., metabolic) functions to continue such that they can continue to swim. This seems to us to be a perfect example of how “structural modification of the environment can power a physiological function outside the body” (turner 2000: 41). in other words, individ-ual Chlamydomonas cells are not operation-ally closed.

« 30 » our goal here is not to point out an unnoticed biological phenomenon, but rather to highlight a mismatch be-tween well-understood phenomena and the enactive framework for their analysis. The assumption that living systems are all characterized by autonomy seems straight-forwardly out of sync with the ways that or-ganisms, and animals in particular, engage with their worlds. This brings out another way of stating our worry: it seems to us that the notion of an ecological niche contra-dicts the idea that organizational closure




Enactivism

241


Enactivism


is always realized by operational closure (perhaps in addition to other operational characteristics). self-maintenance and agent-level stability can only accurately be ascribed to organism-environment sys-tems, we suggest, and not to organisms in themselves. Granted that organisms do display organizational circularity when this is mapped out on a suitably abstract level, and that epithelial and other bodily boundaries have functional importance, neither operational nor bodily bounds do the job of separating that which constitutes an organism from that which does not, and what actually corresponds to this former category will vary depending on the organ-ism’s ecology. This is our first disagreement with enactivism.

Sense-making« 31 » a closely related worry concerns

the logic of sense-making and intrinsic normativity. Enactivism construes organ-ism-environment relations as being “always and everywhere” (to borrow a phrase from Hutto and Myin) a matter of coupling be-tween an individual organism and its cur-rent environment that is modulated by the organism with respect to that organism’s intrinsically produced norms (Barandi-aran, di Paolo & rohde 2009: 378; Baran-diaran & Egbert 2014; Thompson 2011: 30–43; Thompson & stapleton 2009: 24f). There may be cases where the logic of this relation applies – the well-worn example of the E. coli bacterium chemotaxing up a sucrose gradient may fit, for instance – but in general it will not, for two reasons. First, what constitutes a “current” environment, and so what counts as organism-environ-ment coupling, depends on what times-cales have been selected as focal for the investigation. “Current” means something like “temporally proximal on a given times-cale,” although the timescale thus indicated

� is usually given only implicitly by selec-tion of experimental apparatus, and so also

� has no necessary relation to the tempo-ral scales specified by the rhythms that characterize the activity of a given or-ganism or lineage (see Cowley & stef-fensen 2015).

a second problem comes from the regu-latory influence that non-bodily systems

and processes can have on (inter-)action. activity is often organized by norms, such as social norms, that are not intrinsic to any operationally closed system, as well as by the properties of material structures used in particular ways. as an example, consider how both problems apply to the way a spider uses vibrations in its web to detect the location of trapped prey (e.g., Barrows 1915; Barth 1985). in such a case, an organism’s action and sense-making are not, strictly speaking, emergent from the coupling of its body to the environment on the timescale of its anatomical movements. The spider’s action-organization depends on previous organism-environment inter-actions, and on its modification of envi-ronmental structures that stretches well beyond the scale of its present bodily sen-sitivity to changes in its immediate milieu.

« 32 » our concern here is that these aspects of organism-environment interac-tions seem in principle to be beyond the reach of current enactivist accounts. take, for instance, this statement of the enactive notion of cognition:

“ Cognition is the regulated sensorimotor cou-pling between a cognitive agent [i.e., an adaptive autonomous system] and its environment, where the regulation is aimed at aspects of the coupling itself so that it constitutes an emergent autono-mous organization in the domains of internal and relational dynamics, without destroying in the process the agency of that agent (though the latter’s scope can be augmented or reduced).” (Froese & di Paolo 2011: 18)

« 33 » in that paper, cognitive agents are defined as autonomous systems that adaptively regulate their couplings to their environment via neural sensorimotor ac-tivity. in the previous section, we suggest-ed that animals with nervous systems are probably never operationally closed, and so are not autonomous systems; here, we wish to add that the dynamics and the phenom-enal experience of organism-environment coupling cannot be accounted for in terms of normativity that arises from self-orga-nization on the scales of bodily movement and physiological processes.11 This is both

11 | We take it that enactivist definitions of autonomy, adaptivity, etc. can only be understood

because agency involves multi-scalar orga-nization of activity and because the control of activity tends to be widely distributed across bodily and non-bodily processes.

« 34 » The difficulties are clear when we consider an ant navigating by following a scent trail (Gordon 2010). The ant’s body is responsible for fixing the dynamics of its stepping behavior as a form of coupling to the surfaces in its environment, but it can only be said to be carrying out action–per-ception cycles with respect to “goals” – that is, norms – specified by the prior environ-ment-structuring activity of other ants. The ant is moving towards a food source, not engaging in undirected taxis or in mere line-following. it undertakes this action by allowing its small-scale activity to be constrained by the scent trail, which is a slow-changing physical pattern created by conspecifics in order to enable this par-ticular large-scale (nest-to-food) pattern of movement. The ant’s activity is thus guided by a norm of food-finding whose instan-tiation is literally distributed across many members of the ant colony, the food source itself, and the scent trail.

« 35 » or consider norms of attending a classroom environment. Like the ants, teachers make use of built spaces by fac-ing chairs and tables in a common direc-tion and placing displays at eye height in the middle of the wall. They also rely on habits (e.g., hand-raising) and practices (e.g., note-taking) that allow students to participate in classroom events without disrupting the dynamics of the many-to-one focus on the front of the room. There is clearly normativity here – there is a physi-cal constraint on the activity of the agents in the room – but it does not arise as a con-sequence of the operational closure of any specifiable autonomous system. There is no system whose integrity or viability depends on the maintenance of group attention, but this is precisely what the norm calls for, and it is deviations from this to which agents in the room are sensitive. Contrary to enac-tive accounts of the nature of normative ac-tion (e.g., Barandiaran & Egbert 2014), ac-tivity is regulated and organized according

as claims about coupling on these scales, and so this is meant as a general claim about enactivism. see steiner and stewart (2009) for a similar worry.



Cogn

itiv

e SC

ienC

e Co

nCep

tS in

ena

Ctiv

iSm

242


Enactivism



Enactivism

to historical norms, a situation that cannot be readily analyzed using any available nat-uralistic terminology. The important thing to note here is that if we restrict ourselves to the conceptual apparatus of enactivism, we are unable to clearly state what happens when a teacher helps a student calm down, states a fact, or guides a student through the method for solving a problem. nor can we differentiate these events from one an-other or link their occurrence or dynamics to particular mechanisms and processes.

« 36 » We ought not to be surprised by this mismatch between situated action and the enactive framework for understanding it. sense-making and intrinsic normativity must be understood as part of an effort to provide general dynamic-systems-theoretic explanations for very broad classes of phe-nomena. along with “adaptive autonomy,” “sense-making” is intended to specify the formal, systems-theoretic characteristics of agency as a general property. Their inad-equacy in accounting for actual biological organization comes from the fact that the strategies living systems use for intention-ally organizing relations with their environ-ments turn out to be hugely variable over time and between individuals. We strongly suspect that there are no systems-theoretic characteristics common to the ecological niches of all lineages of organisms, but at the very least, if there is a universally applicable logic to organism-environment relations, it is not the logic of sense-making and intrin-sically produced norms. This is our second disagreement with enactivism.

interactivity« 37 » our own approach fits into the

conceptual space opened up by the forego-ing negative comments. it begins with our Maturana-derived view of living systems as self-sustaining but operationally open, and as realizing their molecular self-creation in a wide variety of ways, which constitutively involve environmental structures, con-specifics, other lineages of organisms, and routines, habits, and other ways of stabiliz-ing activity. More precisely, we assume that lineages and individual organisms are con-stituted by the way they interact with the environment. The same assumption applies to intentionally organized activity – this, too, is constituted interactively. That is, not

all living systems display the characteristics of varelian autonomy, and intrinsic teleol-ogy (as enactivists conceptualize it) is not always the basis of how organisms make sense of the world. our approach, based as it is on interactivity and on related concepts such as distributed cognition, is an attempt to come to conceptual and methodological grips with the consequences of these as-sumptions.

Human interactivity« 38 » “interactivity” has always been

used, thus far, to pick out social, cultural, and technological influences on the dynam-ics of human activity. steffensen (2013) of-fers a more concrete definition: interactiv-ity is human activity that is so organized. specifically, interactivity is “sense-saturated coordination that contributes to human action(s),” which can be cashed out as fol-lows.

« 39 » “Coordination” is coupling on the timescales of a body’s molecular pro-cesses, movement of its anatomy, and per-haps other closely connected scales defined by a person’s own modes of self-organizing (Cowley & steffensen 2015). For instance, n. J. Enfield (2013, 2014) describes conver-sation as playing out on an “enchronic” tim-escale, which is “grounded in trajectories of co-relevant actions” (Enfield 2011: 287) rather than in any strictly temporal unit.

« 40 » “sense-saturation” indicates sev-eral things at once. First, it indicates that coordination is experientially and emo-tionally rich, and that both of these aspects are fundamental to how human activity is organized. second, it indicates how human activity is constrained and sometimes con-stituted by the practices and social roles, techniques and tools, articulatory gestures and other patterns of vocal activity, ways of seeing pictures and other aesthetic objects, concepts, and many related phenomena that make up the cultural ecology in which humans act (Hutchins 2014; steffensen & Fill 2014; steffensen & Pedersen 2014; uryu, steffensen & Kramsch 2013).

« 41 » “actions” are directed changes in human-environment systems, where “directedness” is a matter of emergent functions serving to temporarily organize (“soft assemble”) multiple heterogeneous components of a person-environment sys-

tem (Kello & van orden 2009; cf. Bingham 1988; Wilson & Golonka 2013). This view of directedness is similar to scott Kelso’s (1995: 137–149) description of intention-ality as constraints on one or more of the order parameters characterizing a system’s trajectory of change at a given moment. Consequently, it is analogous to – but far more substrate-heterogeneous than – the enactive notion of intrinsic normativity. Because human action is usually co-action with other humans, and because it is con-stituted on multiple timescales at once, the intentionality that “directs” a movement as it begins will be only one of the many regu-latory influences that constrain the agent-environment system as the movement plays out.

« 42 » some of the consequences of such a view can be drawn out with an ex-ample from an experiment on problem-solving, borrowed from steffensen, vallée-tourangeau and vallée-tourangeau (2015; see also Cowley & nash 2013; Kirsh 2015; steffensen 2015). Participants were asked to sort 17 small animal figures into four “hold-ing pens,” which they were to make out of pipe cleaners, in such a way that there were an odd number of animals in each pen. This is known as an “insight problem,” because it requires the subject to change an assump-tion implied, but not stated, by the way the problem is posed. in this case they have to realize that the four pens can overlap, allow-ing an animal to be in two pens at once (if the pens are kept separate, the problem is impossible). one participant in particular displayed an interesting pattern of behavior: she became stymied by the evident impos-sibility of the problem, and, after a pause, began straightening out her pipe cleaner pens, making them into rounder, more per-fect circles. as she did so, the pipe cleaners in her hand caught on the pipe cleaners of the pen sitting on the table nearby, drag-ging it along with her hand movements so that it was partly underneath the one she was holding. When she finished re-shaping it, she set that pen down partly on top of the lower one, and, when she reached out to separate them, she suddenly stopped, brought her hand back to her face, and instead reached out to make the final two pens overlap as well. she went on to solve the problem using overlapping pens.




Enactivism

243


Enactivism


« 43 » steffensen and colleagues (2015) analyze her problem-solving process in this way. in abstract terms, the participant shifts from viewing the problem arithmetically to viewing it set-theoretically, but in concrete terms, her “insight” consists of:

� the velcro-like properties of pipe clean-ers;

� her choice to engage in non-task-rele-vant behavior on the basis of her aes-thetic sensibilities; and

� her ability to notice an accidental over-lap as a (potential) change in the layout of the conceptual space of the problem, in addition to being a previously unrec-ognized material affordance.

This all depends entirely on lived experience as a means of integrating constraints across timescales: the directedness of her actions, initially specified by the bodily act of trying to solve a verbally stated problem, changes as her activity incorporates the pipe-cleaners and animals, and then again when, on the basis of her aesthetic and conceptual abilities, she modifies the sensorimotor affordances of her material situation in accordance with her new understanding that the “animal enclosures” can overlap. in general terms, “problem solving” in this instance consists of an agent making use of the experience of shapes being more or less circular, a specific understanding of animal pens, fences, and fields, and practices for behaving in labora-tories and for solving math problems, in or-der to change the sensorimotor affordances of her unchanged physical setting. Moreover, she does all this according to a conceptually-defined motive that is frustrated by her initial coupling to her material surroundings.

« 44 » The concept of human interac-tivity is intended to pick out precisely this phenomenon of drawing on heterogeneous resources for organizing activity. it concerns how humans use lived experience as a means of integrating local physical constraints with social and technical sources of normativity in order to engage effectively with their en-vironments in ways that change over time.

Building towards a general concept of interactivity« 45 » interactivity is not an explanatory

concept after the fashion of “autonomy” or “sense-making.” it is more like a theoretical opening move: it is a construal of biological

activity that emphasizes a particular set of research interests. in what follows, we will develop both of these aspects of the idea.

« 46 » it is helpful to begin with expert tool use as an illustrative example. This relies on a particular form of bodily and attentional expertise wherein the tool-user attentionally selects specific parameters of the coupled sys-tem consisting of their body, the tool, the ma-terial, and the environment, and then closely monitors and regulates these parameters as the activity plays out (Baber 2015; Baber, Parekh & Cengiz 2014). The skill consists of an ability to use lived experience to manage the coupling of one’s body to a material in the environment by means of a tool, and in service of aesthetic and conceptual goals (cf. Malafouris 2015). This is difficult to concep-tualize in enactivist terms, but easy to con-ceptualize in terms of interactivity. We need only point out that the process of enskillment is not in the manner of a one-and-done event (or a series of such), but is instead a lifelong pattern of body-tool-material encounters, which are connected into a coherent long-term pattern by the experience and bodily organization of the tool-user.

« 47 » interactivity is the idea that ac-tion by biological agents is always funda-mentally heteroscalar (i.e., its dynamics are regulated by processes on several different timescales) and so is strictly non-localizable in space and time (steffensen 2015; stef-fensen & Cowley 2010). That is, organisms are constituted by processes that play out on multiple timescales, and it follows that their actions are organized on multiple timescales as well. if we assume actions are trajectories in some actual spatiotemporal possibility space, then their beginnings, ends, and so on are determined on the appropriate timescale (e.g., the timescale of bodily movement, or of a cascade of neurophysiological process-es, or of an experienced event). However, in order to explain, say, why an organism un-dertook one action rather than another, or when an action functions as an instance of a phenomenological type, we need to ask how the action is organized by (i.e., both consti-tuted and enabled/constrained by) patterns, events, and processes relevant to the organ-ism that are concurrent but exist on differ-ent timescales (Lemke 2000; steffensen & Pedersen 2014). We refer to patterns, events, and processes of this kind as “non-local re-

sources,” by which we mean that they are explanatory resources needed to account for non-localizable actions.

« 48 » This leaves us with several ba-sic conceptual tools. in addition to non-local resources, we have interactivity itself, which is the bringing-together of non-local resources with small-scale bodily dynam-ics in the course of non-localizable (inter-) action that they constrain and constitute. in relation to these concepts, “agency” gives a name to the many various and heteroge-neous ways in which organisms achieve this integration of regulative influence and con-stitutive processes into their activity. There is no consistent, general logic to agency across species, as each lineage will manage their ecological niches in their own ways. These will range from massively heteros-calar in the human case to monoscalar (or close enough as makes no matter) for some bacteria12 (Johannessen & steffensen, forth-coming). an important caveat here is that for humans, interactivity always depends on lived experience (including attention and emotion). other lineages of organisms manage the complexity and non-linearity of their environmental interactions differently, and whether non-locality and interactivity are useful tools for their analysis is a topic for another paper.

« 49 » Consider the difference between speech and the flocking of birds. speech (as opposed to non-speech vocalization) depends on at least two types of non-local resources. The first are population-level statistical patterns of sound units, such as prosodic units and familiar phrases (see e.g., Port 2010a; ramscar & Port 2016). The sec-ond are articulatory gestures (Fowler 2014; Goldstein & Fowler 2003), highly regularized synergies among the various components of the articulatory tract whose repetition re-sults in auditorily, visually, kinesthetically, and acoustically similar events. Both types of resources function normatively in that adult humans regulate their micro-scale vocaliza-tions to match the appropriate patterns and instantiate the appropriate synergies. They achieve this by means of rich phonological memory, entrenchment, sensitization, and related mechanisms, all of which implicate

12 | “Extended metabolism” by Johannessen & steffensen (forthcoming).



Cogn

itiv

e SC

ienC

e Co

nCep

tS in

ena

Ctiv

iSm

244


Enactivism


enaction, and its Relation to Science in an objective Key Fred Cummins

Enactivism

lived experience (Cowley 2016; Cummins 2009; Ellis & Larsen-Freeman 2006; Mac-Whinney 2005; Merker & okanoya 2007; Port 2010b; rączaszek-Leonardi 2014). By contrast, schooling and flocking behavior in birds and fish, while they do depend on emergent population-level statistical pat-terns in body movement, are not obviously amenable to analysis as individual behaviors that draw on extra-individual resources, and it is unclear what it would mean to speak of lived experience in this case (attanasi et al. 2014; Cavagna et al. 2010).

« 50 » Many other non-local resourc-es are ubiquitous in cultural ecosystems.

some of these belong to a general category of highly regularized behavioral patterns that are learned socially, whose function as non-local resources is to enable individuals to regulate their inter-bodily coupling in ac-cordance with specific norms. This category includes rituals, ways of displaying and pub-licly feeling emotions (Fischer & Mansted 2008; Griffiths & scarantino 2005; Jensen 2014; cf. Froese & Fuchs 2012: 213f), and techniques for producing and using artifacts (including tools), built spaces, and other slow-changing material structures. agency with respect to these behavioral patterns – that is, the ability to engage with them as

non-local resources – consists of developing bodily expertise guided by highly skillful at-tending, both of which are often developed by means of having one’s activity closely monitored and guided by older conspecifics (Hoppitt & Laland 2013; Malafouris 2013; sterelny 2012).

« 51 » For these and innumerably many other examples of interactivity, the mecha-nisms that underlie them are only beginning to be understood. our goal here has not been to lay out a well-developed framework for their analysis, but rather to show that an approach based on interactivity is a poten-tially fruitful way of guiding investigation.

mattHeW iSaaC HaRveyis a PhD fellow at the Department of Language and Communication at the University of Southern Denmark. His main interests are the relation between distributed and enactive theories of language and the philosophy of agency and techniques. His PhD is focused on replacing representations (that’s the catchphrase) in linguistic theory, both as a concept and as a suite of terms and associated ways of thinking. With that in mind, his immediate work here is exploring the use of agent-based modeling as a research tool in distributed and ecological linguistics, where the aim is to produce emergent phenomena that are recognizably language-like but cannot be traced directly to representational capacities of individual agents. He arrived at SDU in 2015 after a BA in cognitive science at Vassar College and an MPhil in linguistics at Cambridge.

{

RaSmuS gaHRn-anDeRSenis a PhD fellow at the Department of Language and Communication at the University of Southern Denmark. His areas of research include phenomenology, process philosophy, distributed cognition, and theories about social organizing. His PhD project seeks to outline a theory about social organizing based on an epistemologically clarified phenomenological outset as well as a critique of the analytical concept of context.

{

Sune voRK SteFFenSenis an Associate Professor, and Director of the Centre for Human Interactivity at the University of Southern Denmark (Odense Campus). In 2008 he took the initiative of forming a research group that in 2012 became the current Centre for Human Interactivity. His field of research includes ecological linguistics, distributed cognition, and human interactivity. For more than a decade he has contributed to the development of ecological linguistics through numerous publications, including joint work with Claire Kramsch. His current research combines interaction analysis, ecological linguistics, and situated, distributed, and systemic approaches in cognitive science. He is one of the pioneers of cognitive event analysis, i.e., the study of how short-scale interbodily dynamics, constrained by large-scale, evolutionary, developmental, and sociocultural patterns, enable agents and systems to achieve results. His main empirical interest is interactivity in organizational settings (primarily within the health sector), e.g., expertise, decision making, and problem solving in complex sociocultural environments. Sune is treasurer of the International Society for Interactivity, Language, and Cognition.

{




Enactivism

245


Enactivism


Conclusion« 52 » The relation between interactivity

and enaction is a positive one. The approach-es agree to some extent about research inter-ests and methods, and share certain core no-tions. interactivity-based approaches assume that all organisms are autopoietic, but make no claims about the scales on which they real-ize autonomy, or the synchronic operational means by which they carry it out. in fact, the scales on which a lineage of organisms real-izes their autopoiesis is fundamental to their self-definition as a form of life. at least some organisms are constitutively open in their manner of living, such that processes outside their organizational bounds are frequently, repeatedly, reliably, and fundamentally part of how the organisms regulate their activity and engage with the world. Many such organ-isms are metabolically open (i.e., not literally organizationally closed), but even if they are not, their agency is not accurately character-ized by the model of adaptive autonomy, and

their sense-making is fundamentally a matter of how their activity is embedded in multi-scalar processes, patterns, and events that structure their interactions with the world in experientially specific ways.

« 53 » For the sake of clarity, throughout this article we have been at pains to empha-size what we disagree with in the enactive ap-proach. it bears repeating that we are deeply sympathetic to many varieties of enactivism, core varieties included. if we have a com-plaint about enactivism, it is this: it seems simply to assume operational closure rather than demonstrating it (e.g., in multicellular organisms), and also to assume – again with-out reference to observation-based evidence – that the logic of operational closure is the basis for the organization of activity as it actu-ally plays out, despite the difficulty that both of these assumptions seem to contradict the basic biological facts. While autopoiesis is an abstract conceptual reconstruction of molec-ular self-maintenance, and so evidence-based

in that particular sense, the subsequent no-tions of autonomy and sense-making are not (as indicated by, e.g., di Paolo 2009a, 2009b). Beyond that specific criticism, this essay has tried to show that any phenomenon that is amenable to enactivist explanations is also amenable to explanation using interactivity and associated concepts, where the differ-ence between the two is largely that the lat-ter seeks to explain activity by means of spe-cific non-local resources and mechanisms by which organisms dynamically engage with them, rather than by facts about a-temporal organization. aside from these differences, enactive and interactivity-based accounts will be closely aligned, and it is our hope that cross-talk between them will become easier as interactivity-based approaches are applied in greater detail to questions of mechanism, and as enactive approaches are applied to cul-tural phenomena.

received: 21 July 2015 accepted: 12 January 2016

enaction, and its Relation to Science in an objective KeyFred CumminsUniversity College Dublin, Ireland fred.cummins/at/ucd.ie

> upshot • Enaction, as a paradigm, is still negotiating its position with respect to science done in an objective key. Some of the problems identified by the authors arise by treating enactive descriptions as if they were realist accounts. Negotiating a resolution here will demand progress all round.

« 1 » Enactivism is by now a broad field, broad and inhomogeneous enough

i suspect to accommodate many of the concerns that Matthew Harvey, rasmus Gahrn-andersen, and sune steffensen (hereafter, HGs) present in the target ar-ticle. But the issues of the spatio-temporal distribution of the dynamics of sense-making highlight an important and unre-solved issue. HGs contend that enaction describes agent-environment relations as “in-the-moment coupling” (§2), whereas the activity of humans speaks instead of the integration of heterogeneous resources in organizing their activity (§44). agen-tive activity is non-local in space and time (§47), and so cannot be accommodated within an account of agents as operation-ally closed, and hence localisable in space and time. This seems to HGs to preclude any adequate account of the ecological

transformation that languaging and human sociality bring about (§2).

« 2 » it is surely uncontroversial to assert that human activity must be understood in many different ways, and any given activity may demand acknowledgement of processes at diverse temporal and spatial scales. This is the case, not only for generic “languaging,” but for all activity mediated by texts, elec-tronic media, distributed social institutions, and the countless accretions of technologi-cally enabled, coordinated human living. is this grounds for adopting a completely different approach to agency, treating it as necessarily distributed, rather than a prop-erty of a specific form of self-generating and self-sustaining autonomous organisation? i suspect not, but the reason is rather impor-tant, and my response is not to reject such a

open peer Commentarieson matthew Harvey et al.’s “interactivity and enaction in Human Cognition”



Cogn

itiv

e SC

ienC

e Co

nCep

tS in

ena

Ctiv

iSm

246


Enactivism


Conceptual Foundations of enaction tom Froese

notion of distributed agency either. in much of the present text, HGs treat enaction (in the tradition after Francisco varela) as if it were part of an objective science of biology, concerned only with individuated discrete organisms, but the metaphysical commit-ments of enaction lie elsewhere.

« 3 » to draw the distinction most clearly, it will help to contrast the manner in which creation is treated in Buddhist and in Christian cosmologies. Within the Western intellectual tradition, Christianity has been the unacknowledged background against which all the major moves in phi-losophy of mind and the development of physical cosmology have played out. These have come to adopt a view of time as linear, extending from a single moment of crea-tion (or the singularity antecedent to the big bang, which, being a singularity, cannot itself be considered an event or a moment), and connecting one second to the next. all being is tied to this linear progression, and it is within this mind-independent linear se-ries that life evolved, and that we all progress together from past to future. “Lived experi-ence,” to which HGs repeatedly appeal as a basis for knowledge, is something that either exists, or not, within this framework, as this framework aspires to treat of all that exists. This view of time has gone into the literature as John Ellis Mctaggart’s B-series, or linear time (Mctaggart 1908). But such a universe, as understood within an objective key, is mindless. This is the Copernican-Galilean-newtonian-Cartesian world view tacitly accepted by most researchers, and it has no place for mind (hence the “hard problem” that arises if we adopt that framework). in an objective key, the world comes first, and experience is a puzzle to be understood as something within the world (though never found). appeals to “lived experience” will thus run into opposition if they feature in scientific discourse conducted in an objec-tive key.

« 4 » The 1991 text, The Embodied Mind, sought to inject some Buddhist sen-sibility into discussion of mind and experi-ence. Within a Buddhist framework, sub-ject and world arise together. varela often used the phrase “bringing forth a world” to capture this perspective, and it is linked to the core concept of dependent origination (pratityasamutpada). on this view, experi-

ence is not something found in the world; it is the ground from which both subject and world arise in the always-present now. This is roughly aligned with Mctaggart’s a-series, or lived time, from a phenomeno-logical point of view (especially Heidegger). Contrary to popular opinion, it seems to me that there is no mapping from the a-series to the B-series, or, to put it in contemporary psychological terms, there is no unification possible of the subjective experience of time as flowing (Heraclitian) and the objectivist picture of linear time (Parmenidean). in a strongly constructivist, enactive key, experi-ence is prior to any distinction between sub-ject and world. it is thus the world that needs to be topicalised, not “lived experience.”

« 5 » That these two radically different ontological positions are being confused is clear when HGs speak of an organism’s experience as “being coupled to the world” (§13, emphasis mine). if subject and world co-arise, then the “world” of which we speak is not something from the B-series, some-thing that exists in insolent opposition to the subject. it is rather the milieu, or with some degree of caution, we might use the term um-welt. it is “a” world, arising as complement to “a” subject, and at this point we are clearly being tripped up by language that oscillates wildly between different sets of ontological commitments (or between matters of “orga-nization” and matters of “operation,” to align this with the vocabulary of HGs). This is no-body’s fault. our language is not separable from the worldviews of language users, and the objective key has had a near-monopoly within the realm of scientific discourse.

« 6 » The enactive vocabulary has two main concepts for dealing with the historical antecedents that condition the present, and with the apparent problem of discrete indi-vidual Cartesian minds. These are structural coupling and participatory sense-making, respectively. The former is a formal means by which the constitution of a given system (and hence also its umwelt) can be tied to its historical trajectory and legacy of past interactions (for organisms, this accommo-dates both phylogenesis and ontogenesis). The latter provides a way to acknowledge how the sense-making activities of all life forms give rise to different kinds of mutual dynamic entanglement, and hence to the arising of largely shared perspectives, largely

shared worlds. These concepts are still works in the making, and the distinction they ex-hibit with respect to the treatment of time is a pointer to the linguistic tangles we face as this account matures. They probably need extension if they are to serve in some of the discussions HGs wish to have.

« 7 » Enaction will continue to have a contested relation to the practice of science in an objective key. too small an intellec-tual tradition to influence most scientific practice, it does offer a theatre in which the negotiation of conflicting perspectives can be conducted. it opens a door to the nego-tiation of consensus-based accounts when conflicting value systems are in play, and when consideration of more than one lo-cus of agency is necessary. This is, and will remain, a form of description that will oc-casion self-reflection, that will necessitate an awareness of the framing assumptions of the observer/discussants, and that leads to claims of bounded applicability. it also al-lows discussion of many kinds of subjects, and many kinds of agency (Cummins 2013). to restrict the accounts of enaction to dis-crete organisms alone would be unfortunate. The strengths of the enactive approach, i suspect, lie precisely in its flexibility in ac-commodating agency at many different lev-els, treating subjectivities that are collective, transient, ad hoc, as well as those rooted in individual organisms (or human bodies). it is well positioned to contribute to pluralist ontologies (Latour 2013), and this capacity will only become more important as our discourse extends to meet the demands of a biosphere in crisis, a non-unified species of language users, and the inevitable collision of science with political and religious con-cerns.

Fred Cummins conducts interdisciplinary research into the business of joint (unison) speaking, as found

in practices of protest and prayer. This topic raises issues of collective intentionality that seem to be

best addressed within an enactive framework. It also spotlights such rewarding topics as rhythm, speech-

gesture and speech-music relations, synchronized action, and the aesthetics of ritual practices. He works

at University College Dublin, where he co-directs a postgraduate cognitive science programme.

received: 10 February 2016 accepted: 19 February 2016



247

Conceptual Foundations of enaction tom Froese

Enactivism


interactivity Should aim to extend, not Reject, the Conceptual Foundations of enactionTom FroeseNational Autonomous University of Mexico • t.froese/at/gmail.com

> upshot • Enaction is a diverse research program and some of its texts can be in-terpreted in terms of a critical contrast to interactivity. Yet much of the former has already started to move in a direction favored by the latter: toward systematic studies of how human activity is shaped by social, cultural, and technological in-fluences. Interactivity could therefore help enaction to provide a better account of such highly mediated and augmented forms of sense-making.

« 1 » according to Matthew Harvey, rasmus Gahrn-andersen and sune stef-fensen, the relationship between interactiv-ity and enaction is a “positive one” (§52), with sufficient potential for increased align-ment such that they conclude their target ar-ticle by expressing hope for easier cross-talk (§53). Yet at the same time, they are at pains to highlight shortcomings of the enactive approach, which to them place in-principle limitations on its applicability to the kind of phenomena that interactivity has been most interested in describing in detail, namely the way in which the dynamics of human activ-ity are shaped by social, cultural, and tech-nological influences (§38).

« 2 » indeed, it is no secret that the en-active approach has long been faced with the challenge of overcoming a cognitive gap, i.e., the problem of how to extend its conceptual framework from basic organismic minds to specifically human forms of cognition and action (Froese & di Paolo 2009), and the target article resonates with this criti-cal self-assessment. on the other hand, in recent years there have been an increasing number of applications to precisely these phenomena. This is not the place to review these studies in detail; i will just highlight that they include one of interactivity’s fa-vorite topics: languaging (e.g., Cuffari 2014).

Thus, it is unclear if there are genuine differ-ences between these approaches, or if they can rather be taken as complementary.

« 3 » i agree with Harvey et al.’s conclu-sion that the latter is more likely the case. For example, interactivity is good at producing detailed studies of specifically human activi-ty, which result in descriptions of phenome-na that can be suitable targets for developing the basic explanatory concepts of enaction. Yet this possibility for future collaboration is obscured by misunderstandings and an overly restrictive interpretation of the en-active approach, which i aim to help clear up with this commentary. i hope to show that there is nothing about its conceptual foundations that stands in contradiction to interactivity. neither are they set in stone, and interactivity is welcome to extend them as might be required for accounting for spe-cialized kinds of human activity.

« 4 » Harvey et al. introduce the key concepts of the enactive approach by con-trasting it with Humberto Maturana’s biolo-gy of cognition. This is indeed a useful strat-egy because the former is in significant ways derived from the latter, but unfortunately they portray the development of ideas the other way around. They attribute to Matura-na the position that “circular molecular or-ganization might, without contradiction, be realized operationally by non-circular, con-stitutively open and environment-involving, massively distributed and multi-scalar pro-cesses” (§17). to be fair, Maturana’s writings are notoriously difficult to understand and his theory has subtly changed over the de-cades, and therefore lends itself to conflict-ing interpretations. However, on this point, Harvey et al. appear to be mistaken. to see why, we only need to recall Maturana’s so-called doctrine of non-intersecting phe-nomenal domains. it is worth quoting him at length since this will help to clear up sev-eral issues:

“ autopoiesis describes the internal dynamics that constitutes a living system as a living system in the molecular domain, but a living system also exists as a totality in a relational space where it op-erates as an organism. The constitution of living systems as autopoietic systems entails their con-stitution as organisms as a result of the constitu-tion of their operational boundaries which sepa-rate the molecules that dynamically participate in

their autopoiesis from those that do not. so, living systems exist in two non-intersecting domains, the domain of their components as molecular autopoietic systems, and in the domain in which they operate as organisms (totalities) in a medi-um that makes them possible. These two domains do not intersect, the processes that take place in one cannot be reduced to the processes that take place in the other. Yet, these two phenomenal do-mains modulate each other through the structural changes that take place.” (Maturana 2002: 14f)

« 5 » We can see why Harvey et al.’s ex-ample that “eating and breathing are ways of maintaining organizational closure on the molecular level” (§17) is not compat-ible with Maturana’s theory of autopoiesis. First, eating and breathing are behaviors that take place in the relational domain of the organism and its medium, and as such they cannot be taken as constitutive of the organism’s molecular processes, given that these two non-intersecting domains are only permitted to modulate each other. second, contrary to their interpretation that Matura-na’s work “makes no particular claims about the operational dynamics of living systems” (§17), there are strict requirements. For ex-ample, these dynamics are molecular, they constitute an operational boundary in that domain, and they are internal to that bound-ary. Eating and breathing satisfy none of these requirements.

« 6 » if this is on the right track, then on Maturana’s view, eating and breathing are behaviors that are correlated with the constitutive (molecular) dynamics of the living system, and they can even modulate each other. But their interdependence is fundamentally restricted: the only thing we can say is that they just happen to be cor-related in this way because of a history of natural and ontogenetic structural drift, and because otherwise the autopoietic system in the molecular domain would not have per-sisted. Eating and breathing happen to en-able an organism’s continued survival, but their realization is not part of its identity as the kind of being it is. in other words, if even eating and breathing – behaviors that are necessary for the very survival of the ani-mal – cannot be meaningfully incorporated into its constitutive dynamics, then this aim becomes simply inconceivable for the high-er-level behaviors that interactivity is most



Cogn

itiv

e SC

ienC

e Co

nCep

tS in

ena

Ctiv

iSm

248


interested in, for example a teacher main-taining group attention in the classroom (§35). Here i am assuming that interactivity is interested in explaining such activity in a more satisfying way than by pointing to the contingent conservation of adaptation of the organisms participating in it.

« 7 » autopoietic enactivism can be seen as an attempt to overcome these limi-tations in a gradual and systematic manner. Most importantly, it preserves the claim that an autopoietic system can be described from the perspective of two distinct and mutually irreducible domains, i.e., component proc-esses and interactive processes, but rejects the doctrine of non-intersection. This move, taken together with Francisco varela’s re-definition of autonomy as a form of autopoi-etic self-production that is domain general,1 opens up the possibility for distributed and environment-involving behaviors to play a constitutive role in maintaining organiza-tional closure on the molecular level.

« 8 » Furthermore, behaviors are no longer seen as nothing but contingent out-comes of a system’s conservation of adap-tation and organization, like a sailing boat whose drifting is passively determined by a combination of its physical structure and the prevailing winds (see Maturana’s 1988 concept of ontogenetic structural drift). The concept of an active process of regulation of agent–environment coupling, for example expressed in terms of adaptivity (di Paolo 2005, 2009a), is what enables us to con-ceive of behaviors as organized according to norms in the first place – an idea that has no place in Maturana’s anti-teleological frame-work (see villalobos and Ward, this issue). The upshot of these conceptual changes is that the enactive approach is in a position to address precisely the kind of phenomena in-teractivity finds most compelling: constitu-

1 | as part of the extension from autopoiesis to a domain-general concept of autonomy, varela (1979) introduces the term organizational closure. Confusingly, Harvey et al. attribute this concept to Maturana’s autopoeitic theory, although he has never used this term, while limiting varela’s theo-ry to operational closure (§19). it would be good if Harvey et al. gave clear definitions of these terms, as they put a lot of weight on their being different although many scholars use them inconsistently or even interchangeably.

tive dynamics that incorporate normatively organized behaviors that are distributed and environment-involving, and that can serve as a bridge from life to society (e.g., di Paolo 2009b).

« 9 » some illustrative examples may help. First, certain behaviors can now be seen as directly playing a constitutive role in the maintenance of organizational clo-sure. For example, movement can enable a self-producing chemical network to persist in concentration ranges that would other-wise cause it to disintegrate (Froese, virgo & ikegami 2014). second, the chemical proc-esses of autopoiesis no longer need to be seen as only internal to an organism’s body. For instance, when a worm excretes a sticky substance into its tunnel and thereby ex-ternalizes a part of its digestive process, we should take the worm’s constitutive chemi-cal dynamics to be distributed across the body–environment system (virgo, Egbert & Froese 2011), forming what we might call an extended metabolism. Third, the distributed constitutive dynamics can also be environ-ment-involving. For example, Ezequiel di Paolo (2009a) has appealed to the fact that several species of insects are able to breathe underwater by trapping air bubbles in their hairs, mediating the regulation of environ-mental coupling via the incorporation of an external resource, and thereby bringing forth a more subaquatic way of life.

« 10 » These examples should go some way toward addressing the concern ex-pressed by Harvey et al. that non-bodily as-pects of organism-environment interactions seem in principle to be beyond the reach of current enactivist accounts (§32). But can enaction also account for the other-related normativity of social interactions? Harvey et al. are skeptical and cite Froese and di Paolo’s (2011) definition of cognition in sup-port of this concern (§§32f). However, they neglect to mention that we emphasized that the domain of cognition should be thought of in terms of a hierarchical decoupling from the metabolic domain, thereby permitting cognitive processes to be organized by non-metabolic norms. once it has been shown how non-metabolically mediated regulation of coupling can be a part of the constitutive dynamics of the organism, for example in terms of recurrent neural sensorimotor ac-tivity and habits, nothing stands in the way

of addressing phenomena whose forms of mediation are increasingly distant from the spatiotemporal scales of an individual’s bio-logical embodiment, i.e., what Harvey et al. refer to as non-local resources.

« 11 » indeed, one of the most exciting contributions of the enactive approach in the last decade has been its investigation of how interactions among individuals can co-determine their behavior and their organi-zation, which has resulted in the rejection of one of the essential premises of traditional cognitive science, namely methodological individualism. For example, when two or more individuals are co-regulating their in-teraction dynamics, and thereby their sense-making activity, they can be said to be shap-ing each other’s experience, as captured by the concept of participatory sense-making (de Jaegher & di Paolo 2007). This interac-tion process can become self-sustaining, as each action tends to evoke a re-action, and we thus have a case of varela’s generalized concept of autonomy in the domain of social interactions. it permits us to study individu-al behaviors and social interaction dynamics in terms of two interdependent yet mutually irreducible domains.

« 12 » The concepts of participatory sense-making and the autonomy of the in-teraction process have been criticized by proponents of interactivity, who claim that these concepts are insufficient for account-ing for most human social activity because (a) they neglect the role of the impersonal resources of culture, and (b) they fail to ap-preciate the role of lived experience in devel-opment (Cowley & Gahrn-andersen 2015). it is not clear to what extent Harvey and col-leagues share the second concern; they rec-ognize that first-person experience has been one of the cornerstones of enaction since the beginning (§6). Certainly, development and phenomenology of sociality have not been neglected topics (Fuchs 2013). However, Harvey and colleagues agree with the first concern, which amounts to the claim that the enactive approach is incapable of ad-dressing the role of non-local sociocultural resources because it relies on the concept of operational closure, both at the level of the individual and of the interaction process.

« 13 » i hope to have shown that this worry is misplaced. it is precisely the ex-tension of varela’s generalized concept of



249

Disagreement Between interactivity and enaction nathaniel F. Barrett

Enactivism


autonomy to the social domain that permits us to develop an account of the influence of heteronomous sociocultural processes on human activity because it prevents any at-tempt at reducing these processes to the ac-tions of the individuals that constitute them at any one point in time (torrance & Fro-ese 2011). in other words, the autonomy of the social world is a necessary condition for claiming that its interaction processes are genuine and have an effect on our lives, and yet they operate on timescales that go be-yond individual actions, and in some cases beyond our lifetimes.

« 14 » to be fair, it might at times sound as if the enactive approach is locked into sub-personal explanations of the here-and-now. But this legacy of Maturana’s biology of cognition, which also emphasizes that the behavior of organisms must be understood on the basis of their being state-determined systems, is simply a way of forcing us to distinguish functional from operational explanations. For example, when we want to explain why an organism undertook one action rather than another (§47), we can appeal to non-local resources, for example the historical norms instituted by the legal system. But this does not suffice as an expla-nation of how, in that very moment, the op-erations underlying the individual’s choice of action were realized in accordance with those cultural norms. The challenge is to provide an account of how such non-local resources shape local human activity, and it seems to me that the enactive approach is well positioned to do so (Gallagher 2013).

« 15 » on this view, phenomenology, in-tersubjectivity, and the products of culture must be analyzed in an interdependent man-ner (Havelange 2010). it is precisely in this interdisciplinary study of human cognition that interactivity could offer much needed help to enaction, while interactivity could benefit from extending, rather than reject-ing, enaction’s key explanatory concepts.

« 16 » For instance, Harvey et al. do have a point that the currently most popu-lar account of sense-making, which tries to ground all forms of meaning only in terms of perturbations to the different kinds of vi-ability associated with an organism’s way of life, seems too limited to account for all fac-ets of our lived experience. to return to an earlier example, what kinds of viabilities are

perturbed during the teacher’s task of regu-lating attention in the classroom? Enaction and interactivity may have to work together to find another way of accounting for such abstract normative organization of behav-iour, one that is hierarchically decoupled from an agent’s basic need to regulate its responses so as to directly mitigate threats against the integrity of its way of life.

tom Froese is Associate Professor at the Applied Mathematics and Systems Research Institute

(IIMAS) of the National Autonomous University of Mexico (UNAM) in Mexico City, where he heads the

4E Cognition Lab. He is also a member of the Centre for the Sciences of Complexity (C3), also at UNAM.


What is at Stake in the Disagreement Between interactivity and enaction?Nathaniel F. BarrettInstitute for Culture and Society, Spain • nbarrett/at/unav.es

> upshot • To sort out their differences with enactive theory, interactivity theo-rists would do better to focus on opera-tional closure only insofar as it consti-tutes a condition of intrinsic normativity or self-regulated coupling.

« 1 » The target article presents a com-parison of two constructivist perspectives in cognitive science – the newly emerging in-teractivity-based approach and the increas-ingly influential enactive theory (especially the post-varela branch) – with the dual pur-pose of articulating the theoretical advan-tages of the former while highlighting, in contradistinction, certain critical limitations of the latter. in my estimation, Matthew isaac Harvey, rasmus Gahrn-andersen and sune vork steffensen succeed in presenting interactivity as a distinctive viewpoint that promises new insights into the nature of cognition, but the central concept of “non-local resources” remains unclear.

« 2 » also, and for related reasons, it remains unclear whether their criticism of enactive theory hits its mark. i have doubts about the accuracy of some of the authors’ claims about enactive theory, but more im-portantly i do not believe that they have shown that the critical issue of operational closure bears directly upon the core explan-atory concepts of post-varela enactive the-ory, especially its understanding of cogni-tive agency as a normatively self-regulated process of structural coupling. Even if op-erational closure has been wrongly assumed by enactivists after Francisco varela, it is questionable whether post-varela enactive theory has much at stake in this assump-tion. if operational openness can be granted by enactivists (and perhaps it already it is), then the authors’ argument loses much of its critical leverage, and it becomes difficult to assess their claim that interactivity offers a substantially different theoretical basis for “4E cognitive science.”

« 3 » in hopes of pinpointing the crux of the authors’ disagreement with enactive theory, i will attempt here to summarize what i see as the key points of their argu-ment. Like the authors, i am sympathetic to many aspects of enactive theory but i also find the post-varela notion of “sense-making” to be highly problematic, albeit for different reasons (Barrett 2015a, 2015b). accordingly, my intent is not simply to de-fend enactive theory against criticism but rather to contribute to the authors’ project of mutual clarification.

« 4 » to begin, the interactivity ap-proach is distinguished by its starting point: it starts with human social interac-tion and languaging and seeks to apply insights gained from this context to the construction of more general theories and models of cognition. it therefore shares with enactive theory a radical thesis about continuity of mind, i.e., that the “higher” forms of human cognition share basic traits with the agency of primitive organisms. Mainstream “cognitivist” cognitive science also frequently assumes language as the paradigmatic form of cognition but usually understands language in rule-based, rep-resentational terms. in contrast, the inter-activity approach asserts that language can be understood in fully nonrepresentational terms, where representations are under-



Cogn

itiv

e SC

ienC

e Co

nCep

tS in

ena

Ctiv

iSm

250

Disagreement Between interactivity and enaction nathaniel F. Barrett

Enactivism


the Role of allostasis in Sense-making Robert Lowe

stood as content-bearing states or struc-tures (§12).

« 5 » if we take up an interactivity-based approach, the authors suggest, we are more readily able to realize the importance of heteroscalar dynamics and distributed fea-tures – what they call “non-local resources” – for understanding cognitive agency of all kinds. They also claim that the assumption of operational closure prevents post-varela enactive theory from accounting for these non-local resources. The concept of non-local resources seems to be both the central insight of the interactivity approach and its greatest expository challenge. in the case of human cognitive agency, the non-local re-sources par excellence are “social, cultural, and technological influences” (§38). These influences evidently entail norms of a his-torical and distributed nature such that they lie outside the “immediate spatiotemporal scope” (§3) of a particular human action. it should be noted that this non-local nature is not just a problem for enactive theory: according to the authors, such influences “cannot be readily analyzed using any avail-able naturalistic terminology” (§35). Be that as it may, while social, cultural and techno-logical influences on human interactivity serve as paradigmatic examples of non-lo-cal resources, the authors wish to show that non-local resources are widespread features of cognitive agency (as demonstrated by such examples as spiders + webs and ants + scent trails).

« 6 » it is difficult to tell whether the concept of non-local resources offers us a new advance in understanding or simply a re-description of features that are already (more or less) understood in different terms. among the things that would seem to count as non-local resources are habits and affor-dances, both of which present challenges to some explanatory frameworks but cer-tainly not to all (for enactive work related to these concepts see, e.g., Barandiaran & di Paolo 2014; Buhrmann, di Paolo & Baran-diaran 2013). at a more general level, the acceptance of non-locality seems to require acknowledgement of the relationality and continuity of natural processes. are features of relationality and continuity actually out-side the scope of “any available naturalistic terminology”? Perhaps after david Hume it became commonplace for philosophers

to deny these features as part of experi-ence, but good arguments are on hand to show the inconsistency of this denial (e.g., Whitehead 1967), and in any case it is doubtful that scientific explanation has ever been faithful to Hume’s strictures. More to the point, if one accepts relationality and continuity as part of our understanding of natural processes – and i see no reason to believe that enactive theory does not – then how is one to define the “immediate spatio-temporal scope” to which enactive theory is supposedly confined?

« 7 » Here is where i suspect the criti-cisms of the authors are misplaced. i see no reason to ascribe to enactive theory any sort of commitment to the “immediate spatio-temporal scope” of organism-environment coupling. Yet this seems to be the upshot of the authors’ argument that enactivists wrongly assume operational closure. For instance, they express the key insight of interactivity – the dependence of human cognition on non-local resources – with the claim that human action evidently “cannot be exhaustively explained in terms of the coupling of bodily movement to surfaces of the environment” (§3). The implication of this statement is that enactive theorists be-lieve that human action can be exhaustively explained in this way, but i see no evidence of this narrow view.

« 8 » Perhaps the authors mean to show that this narrow view of cognition is an unacknowledged consequence of a com-mitment to operational closure. so let us consider operational closure more carefully. The main point of the authors’ historical critique of enactive theory is to show that after varela, the concept of autonomy devel-oped by Humberto Maturana shifted from a relatively open notion of organizational closure to the much more restricted no-tion of operational closure (see §27). Thus, according to the authors, “varelian auton-omy” and operational closure are roughly synonymous, and the most precise defini-tion of the problem with this identification is that it specifies “the defining properties of the system’s actual operational dependen-cies over time” (ibid.), and this specifica-tion of operational dependencies over time precludes heteroscalar coupling or indirect kinds of dependency. so we are back to the claim that enactive theory confines its de-

scription of cognitive agency to a narrow view of organism-environment coupling “in the moment.” against this view, the au-thors argue – using several cases of non-hu-man cognitive agency – that the processes that maintain an organism loop through the environment at various timescales and through various indirect kinds of depen-dency.

« 9 » Based on my reading of post-va-rela enactive theory, however, the essence of “varelian autonomy” is not operational clo-sure per se, but rather the way in which or-ganism-environment coupling is regulated by the organism with respect to its own self-generated norms. What decisively divides post-varela enactive theory from Matura-na, then, is the insistence of the former on the asymmetry of organism-environment coupling (Barandiaran, di Paolo & rohde 2009). This asymmetry – based on some kind of homeostatic self-regulatory capacity – is what distinguishes living systems from other kinds of structural coupling and is the hallmark of cognitive agency qua “sense-making” (e.g., di Paolo 2009a). The authors recognize this notion of “sense-making” as central to post-varela enactive theory (sec-tion 3.2, §§31–36), but again their critique is steadfastly focused on what they see as the narrowly defined timescale of this self-regulatory capacity. Their insistence on this point seems excessive – again, nowhere do they provide evidence of a commitment to a single timescale of coupling – but let us leave this question open.

« 10 » Here is where i believe the ac-tual crux of the disagreement between in-teractivity and enactive theory should lie, if indeed there is a serious disagreement to be had. to my mind, the most important question is whether the kind of operational openness insisted upon by the authors – heteroscalar coupling and all the “non-lo-cal resources” to which it gives access – is compatible with the self-regulated nature of coupling that defines cognitive agency for post-varela enactivists. in other words, can organism-environment coupling be both heteroscalar and self-regulated? i do not see why not. But more importantly, are the au-thors arguing that it cannot? it is not clear.

« 11 » The authors’ main objection to the intrinsic normativity of sense-making is that many of the norms that influence be-



251


Enactivism


haviour are not intrinsic to the individual organism. But enactivists could acknowl-edge this point and still insist on intrinsi-cally generated normative regulation as the mark of cognitive agency. That is, it might be possible to have intrinsic normativity without excluding other kinds. in any case, i am suggesting that enactive theory only has a stake in operational closure insofar as it is necessary for self-regulated coupling, and their claim about self-regulated coupling is that it is necessary for cognitive agency – but not that it is exhaustively explanatory of all aspects of all cognitive behaviour.

« 12 » in certain places, the authors do seem to reject intrinsic normativity as a necessary condition for cognitive agency (e.g., §37). But again their position is not sufficiently clear, as they might only mean to say that intrinsic normativity is not enough. in an important passage (§41), they define action in terms of the temporary organization of heterogeneous components of the person-environment in a way that fails to specify – perhaps deliberately – how the task-specific and goal-directed nature of this organization is decided. More tellingly, when speaking of scott Kelso’s theory of in-tentionality (ibid.; Kelso 1995), they neglect to specify that intentionality refers not just to any constraint on the order parameters of system trajectory, but rather just those con-straints that are self-imposed by the person.

« 13 » Here is a missed opportunity to clarify the crucial distinction between what is necessary for cognitive agency and what is sufficient to explain the actual trajectory of system behaviour. Certainly there are multiple constraints on human action as it plays out, and many of these constraints come from outside the individual – just as the authors say – but what makes an action intentional is not just its original impetus but also the continuing role that the indi-vidual plays in the coherent integration of these constraints so that the action remains directed toward a specific task. in the case of intentional action, therefore, not all con-straints that arise from the total person-environment system are equal: only those that arise from the person are responsible for the selection of relevant environmental constraints (e.g., affordances) and the inte-gration of these constraints into a coherent, task-specific course of action.

« 14 » a closer look at the authors’ ac-count of human interactivity suggests that they also give the person a privileged role in the normative regulation of behaviour. That is, even while they locate multiple sources of normativity outside the person, the per-son remains for them a necessary source of the norms that integrate these and other non-local resources into a unified course of action. This is apparent in their description of a problem-solving experiment in which a subject’s “insight” consists in the re-orga-nization of various materials so as to bring different affordances into play. The acciden-tal redistribution of the materials may have triggered the insight, but only the person in her intentional capacity – as directed by a “conceptually-defined motive” (§43) – can realize that the new configuration provides a solution.

« 15 » The authors’ observation that hu-man action consists of the integration of heterogeneous components across multiple timescales seems to me a valuable insight, deserving of careful consideration. Even so, when we ask what decides or governs this integration, i remain convinced that the enactivists are on the right track with their proposal of intrinsic normativity. on the other hand, there may be serious problems with the enactive concept of normativity (Barrett 2015b), and no doubt the inter-activity approach has much to contribute to this issue. With respect to the example of the problem-solving situation just dis-cussed, i would say that the weakness of enactive theory is not that it cannot account for the use of distributed resources but that its picture of self-generated normativity cannot account for the switch from one way of organizing interaction to another.

nathaniel Barrett is a research fellow and member of the Mind-Brain Group at the Institute for Culture of

Society (Pamplona, Spain). His current research focuses on the nature and role of value in life and mind.


the Role of allostasis in Sense-making: a Better Fit for interactivity than Cybernetic-enactivism?Robert LoweUniversity of Skövde/University of Gothenburg, Sweden robert.lowe/at/his.se

> upshot • In contrasting an interactiv-ity account alternative to variants on the enactive approach, the authors discuss the role of sense-making. They claim that their interactivity perspective, un-like enactive approaches, accounts for a dependency on “non-local” resources characteristic of many organisms. I draw attention to the cybernetic-enactivist perspective on homeostatic sense-mak-ing, which may fundamentally fail to explain the operationally open nature of organismic regulatory processes better captured by Sterling’s notion of “allosta-sis.” This sense-making model appears to be more in line with Harvey et al.’s inter-activity framework, focusing as it does on the regulatory effect of extrinsic “norms” rather than the cybernetic-enactivist fo-cus on intrinsic “norms.”

« 1 » Matthew isaac Harvey and col-leagues, in their target article, present what they consider to be two core concepts of enactivism that are fundamentally flawed, i.e., that (a) (all) organisms are operationally closed, and that (b) “sense-making” depends on engagement with non-local resources. i focus my commentary on sense-making, with particular reference to the notion of homeostasis given by advocates of cybernet-ic-enactivism, and contrast it with that of al-lostasis, as defined by Peter sterling (2004, 2012) and Jay schulkin (2011a, 2011b). i place focus on the latter notion and discuss it with respect to Harvey et al.’s interactivity position.

« 2 » according to Ezequiel di Paolo (2005), in Harvey et al.’s terminology a post-Varela autopoietic enactivist (§23), the origi-nal formulation of autopoiesis (varela 1979; Maturana & varela 1980) did not sufficiently explain sense-making. di Paolo brought in



Cogn

itiv

e SC

ienC

e Co

nCep

tS in

ena

Ctiv

iSm

252


the notion of adaptivity – “the capacity of the organism to regulate itself with respect to the boundaries of its viability” (di Paolo 2005: 429) – in order to bridge this per-ceived gap. The paradigmatic case of such an adaptive, sense-making1 agent, for the post-Varela autopoietic enactivists has been the homeostat of ross ashby (1952) – a cyber-netic manifestation of classically conceived biological homeostasis. simply, adaptivity, in the homeostat, is construed as exploit-ing a homeostatic process through first- and second-order feedback. First-order feed-back is internal – the agent senses whether its essential conditions of self-maintenance (essential variables) are compromised. second-order feedback concerns random reconfigurations of parameterizations of sensor-motor couplings whose stability is arrived at by “positive” first-order feedback (the agent is in equilibrium). double order feedback imbues, in this non-autopoietic ab-straction of an organism, adaptivity mecha-nisms of the type that enable sense-making. in a recent publication of anil seth (2014),2 the ultrastable property of ashby’s homeo-stat was again alluded to. in this case, he described the second-order feedback as “al-lostatically [re-organizing] a system’s input-output relations when first-order feedback fails” (seth 2014: 7) and then defines allo-stasis as “the process of achieving homeo-stasis,” originally presented in Xiaosi Gu and Thomas FitzGerald (2014). The use of the term allostasis is, thus, of contemporary rel-

1 | tom Froese and John stewart (2010) point out, however, that ashby’s homeostat, though providing an abstraction useful for popu-larizing enactivism, in lacking chemical self-re-production in far-from-equilibrium conditions, is non-autopoietic, and, therefore, not a biologi-cal “sense-maker.” With this caveat in mind, i contend that ashby’s homeostat provides a useful point of reference for discussing adaptive process-es integral to sense-making mechanisms.

2 | seth suggests his particular predictive processing account of cognition is an example of weak enactivism – incorporating the notion of agent-environment coupling through active perception. seth describes himself as not holding to strong enactivism since he adheres to a frame-work that uses internal models. He might thus be viewed by Harvey et al. as a post-post Varela auto-poietic enactivist.

evance to cybernetic-enactivists. However, there are many biological exponents of al-lostasis holding differing definitions. These definitions allude to different bioregulatory effects and have implications for the role of intrinsic-extrinsic3 norms in sense-making. using the ashbyan homeostat (cybernetic-enactivist) analogy, we summarize three homeostatic-allostatic regulatory positions for sense-making:1 | Reactive homeostatic regulation: sense-

making exploits behavioural corrective mechanisms (second-order feedback) whenever homeostatic disequilibrium is signalled (via first-order feedback);

2 | Transient allostatic regulation: Homeo-static disequilibrium is permitted in the face of ongoing behaviour towards ex-trinsic “goals” or norms. satisfaction of such norms has the effect of facilitating the intrinsic norm of long-term homeo-static equilibrium. The reactive process (Position 1) is embedded within this allostatic process. second-order feed-back involves behavioural corrective mechanisms but may include transient re-setting (autonomic effects) of first-order homeostatic bounds in order to meet present demand.

3 | Non-transient allostatic regulation: a neurophysiological state depends on ex-trinsic demands relevant to survival and reproduction in accordance with “goals” and extrinsic norms. Bounds of opera-tion of certain variables are a constraint on, rather than the point of, sense-mak-ing. second-order feedback involves be-haviour suited to meet present demand and includes redefinition of first-order homeostatic bounds according to ever-shifting extrinsic demands.« 3 » i will now further elaborate on

how these three perspectives on sense-mak-ing map to cybernetic-enactivist thinking, and also relate Position 3 to the interactiv-ity perspective put forward by Harvey et al. in the target article. Whereas the focus for sense-making in Positions 1 and 2 is on (short-term or long-term) adherence to in-

3 | For extrinsic norms, we might also read interactive norms – the critical point is that the norm is the result of the organism’s perceived mode of coupling it has with its (social) environ-ment.

trinsic norms, the focus in Position 3 is on extrinsic, or interactive, norms.

« 4 » Position 1, is most obviously compatible with the post-Varela autopoietic enactivism perspective of di Paolo (2005). sense-making, in this view, is enabled by adaptive mechanisms that help realize au-topoiesis: “adaptivity establishes an inner norm, a distinction between good, neutral, and bad ways of realizing autopoiesis” (di Paolo 2005: 440). The cybernetic-enactivist stance is epitomized by di Paolo’s (2003) light-seeking robot homeostat (see iizuka, ando & Maeda 2013 for an alternative light-seeking robotic homeostat). This artificial organism satisfies an “intrinsic” norm4 (battery maintenance) through reconfig-uring the initially random sensor-motor mappings of the robot whenever the level of this variable veers outside its ideal bounds. This has the effect of inducing the robot to encircle a light source by continuous reac-tive reconfiguring of its sensor-motor map-pings. The equilibrium state of the artificial organism is constituted by the internal and behavioural stability achieved by coupling internal value (maintenance of battery level) to a limited range of sensor-motor mapping re-configurations. Thereby, this mode of cir-cular organism-environment coupling com-prises a type of operationally closed system.

« 5 » Position 2 might be reached by an elaboration of the di Paolo (2003) enactive perspective. in this case, internal instability (homeostatic disequilibrium) may be im-posed on the organism-environment cou-pling as a result of a transient environmental change. However, the homeostat, for the duration of the extrinsically imposed tran-sient, will be in a maladaptive state. in seth’s predictive processing model (2013, 2014), on the other hand, a (predictive) directed-ness of behaviour during such a transient promotes long-term self-maintenance of the organism. This potentially entails top-down suppression or modification of interoceptive prediction errors (deviations from homeo-static bounds) and either autonomic or “al-lostatic” behavioural responses that enable the re-establishment of homeostasis. allo-stasis, here, is perceived as the mechanism

4 | However, “ideal” homeostatic bounds for battery levels are still imposed upon the ro-bot’s design from without.



253


Enactivism


for achieving homeostasis (Gu & FitzGerald 2014; seth 2014). a biological perspective that resonates with the aforementioned is that of John Wingfield (2004). allostasis, compared to reactive homeostasis, provides greater flexibility of the organism-environ-ment coupled system as a whole and entails “emergency adjustments of physiology and behaviour to maintain homeostasis in the face of challenges” (Wingfield 2004: 312).

« 6 » in the case of Position 3, sterling (2004, 2012) has advocated that the reactive homeostasis model is wrong. in sterling’s account, allostatic processes do not sub-serve homeostatic local states. also at odds with Position 2, allostatic processes are not merely “emergency systems superimposed on the homeostatic model” (sterling 2012: 2). The standard medical model of homeo-stasis, the classically conceived Walter Can-non (1929) model of defending set points (though see Craig 2015), insofar as medical treatments focus on localized symptoms, e.g., drugs to reduce blood pressure, is mis-guided. allostasis, here, is viewed as a top-down interactive-constitutive process that recruits resources from many (local) physi-ological systems to meet predicted current demand. This perceived demand is often extrinsic, e.g., set by social norms,5 which exerts a pressure, both in terms of survival and reproduction. sense-making is, thus, constrained, rather than set, by self-main-tenance. Moreover, the ideal physiological state for the organism – in terms of self-maintenance – is not static. sterling uses the example of blood pressure, which fluctuates throughout the day as a consequence of daily rhythms and may be elevated for prolonged periods according to extrinsic demands. one can imagine the human sat at his/her office working long hours to safeguard his/her job being facilitated in this endeavour via elevated blood pressure. rather than viewing this as the ideal state, or a transient

5 | sterling gives the example of cortisol lev-els found in the fecal matter of troops of baboons and that the level found reliably predicts the social status of the baboon. The social hierarchy of the baboon troop is organized around this variable; female baboons, for example, deal with increased cortisol levels (e.g., caused by bereavement) by extending their social network, which serves to reduce cortisol levels.

coping mechanism for facilitating long-term homeostasis, any given physiological state may be viewed as an adaptive response to the particular social pressures that occur within an organism-environment coupling, and are themselves subject to change.

« 7 » sterling’s (2012) discussion of fluctuating blood pressure over cycles of activity and according to changing ex-trinsic demands is perhaps an example of what Harvey et al., in §31, refer to as “the rhythms that characterize the activity of a given organism or lineage.” The timescale of evaluation of sense-making for the organ-ism should encompass daily (and possibly seasonal) cycles on this model’s account (compare to Harvey et al.’s multiple tim-escale view of sense-making in §47). For sterling (2012), a healthy system should be defined as one of “optimal predictive fluc-tuation” (sterling 2012: 9) and inflexibility of essential variables6 regarding their range of operation should be met with therapeutic measures designed to restore flexibility of the response, e.g., that influence the percep-tion of meeting extrinsic demands.

« 8 » From a cybernetics-enactivist perspective, the set boundaries of the ideal homeostatic range that provide the met-ric for first-order feedback are in constant flux according to this allostatic model. in turn, the sensor-motor mappings (the level at which seth ascribes a role for allostasis) are then constrained by the top-down per-ceived extrinsic demands that are required to be met and that in turn modulate the boundary settings of first-order feedback. While Position 2 implies re-establishment of set boundaries – to the range ideal for self-maintenance – following meeting of the extrinsic demands, Position 3 implies an effect on these bounds imposed by the degree of exposure to the particular sets of extrinsic demands. From a cybernet-ics perspective, Position 3 is more usefully conceived as an allostat (see also Muntean & Wright 2007, and vernon et al. 2015 for reference), which regulates according to extrinsic/interactive norms. Position 3 ap-

6 | naturally some essential variables are more flexible regarding their range of viable op-eration than others. Brain glucose levels, for ex-ample, are required to be tightly regulated to en-sure organismic viability.

pears to be consistent with, therefore, the interactivity account of Harvey et al., in the target article. organismic regulation is interactive, according to changing social norms, to such an extent that constitutive organization of the organism “involves body-environment interactions” (§26). The physiological and metabolic activation patterns required to deal with such norms are necessary for organismic viability and comprise “lineage-general ways of main-taining […] molecular self-reproduction” (§26) focused on flexibility according to perceived demand that may vary over days, seasons, and the organism’s lifetime.

« 9 » Extending the definition of allo-stasis given in Position 3, to emphasize fur-ther the links with Harvey et al.’s interactiv-ity perspective, i refer to schulkin’s (2011a, 2011b) term social allostasis. This notion entails “an evolved set of anticipatory mechanisms that facilitated the regulation of the internal milieu amidst an expanding social milieu” (schulkin 2011b: 7). Criti-cal to social allostasis is social competence and maintenance of social status (schulkin 2011a,b). Feedback from the social milieu provides a type of predictive regulation differentially impacting on individuals in relation to perceived extrinisic (interactive) demands, i.e., disruptions of such social status and competence.

« 10 » it has been argued that, for hu-mans, fundamental to the evolution of cog-nitive capacities has been the simultaneous evolution of an adaptive stress response (Cummins 2012). such an adaptive mecha-nism may have been necessary to accom-modate the increasing allostatic load (wear and tear effects of stress) consequent to the expanding goal-directed temporal horizons and complexifying social milieu that con-fronted our early hominin ancestors (Cum-mins 2012). For socially complex organisms such as humans, constitutive regulation of individuals should therefore also be consid-ered in terms of the social context in which the lineage has evolved. Essentially, predic-tive regulation of lower level constitutive pro-cesses is evolutionarily adapted to accommo-date states of both an individual (intrinsically normative) and environmental (extrinsically normative) nature. From the perspective of Position 3, again, intrinsic normativity con-cerns the constraints imposed on the organ-



Cogn

itiv

e SC

ienC

e Co

nCep

tS in

ena

Ctiv

iSm

254


ism for adapting to extrinsic demands. The degree of complexity of social construction on an evolutionary and organismic timescale is such that organisms are required to be op-timally flexible to deal with changing extrin-sic/interactive demands.

« 11 » on the above reading, the allo-static process is constitutive to a socially, and ontogenetically, constructed type of sense-making. Critical to debate is whether sense-

making, so-conceived, can be captured within an extended cybernetic-enactivist framework of the type put forward in this commentary, or whether it is of greater ex-planatory value to adhere to the interactivity framework, as Harvey et al. propose.

Robert Lowe is an Assistant Professor (Docent) in Cognitive Science and Cognitive Robotics and

works at the University of Skövde, at the Interaction

Lab, and University of Gothenburg, where he leads research at the ICE (Interaction, Cognition and

Emotion) Lab. His principle areas of interest concern cognitive-affective architectures for robots, affective

processes in joint action, reinforcement learning models of emotion, and evolutionary robotics

approaches to the study of adaptive behaviour.


authors’ Response explanatory pluralism and precise Conceptual Development Matthew Harvey et al.

Matthew Isaac Harvey, Rasmus Gahrn-Andersen & Sune Vork Steffensen

> upshot • We agree with commenters that enactivism incorporates a broad variety of methodologies, metaphysical stances, concepts, and investigative ap-proaches, and that this is a good thing. However, we remain concerned that au-tonomy and sense-making are problem-atic concepts for post-Varelian enactiv-ism, and that they form the foundations of a conceptual framework that may hamper the development of effective ex-planations for cognitive activity, as well as the paradigmatic aspirations of this particular enactivist approach.

introduction« 1 » We would like to thank all four

commenters for their careful readings and thoughtful responses to our target article. They have advanced the discussion a great deal, and we are very pleased to see that our article has been read as we hoped it would be, and to participate in subsequent conver-sation. This is an ideal opportunity to do as we all agree we ought, and see what can be done to clarify enactivist concepts vis-á-vis interactivity-based approaches, and vice versa. We also want to say up front that we have no qualms about the enactivist label, or about enactivism as a broad collection of

perspectives and research practices. We ap-plaud efforts to take lived experience, Bud-dhist philosophy, and first-person perspec-tives into account in cognitive science. our concerns are narrow but (we hope) quite sharp, and have to do specifically with au-tonomy, operational closure, sense-making, and intrinsic normativity.

« 2 » our response is divided into two parts. First, we discuss the proposals made by the commentaries about an enactive perspective, and whether specific enactive concepts might be modified or adapted in response to our negative argument. it should be noted that, with one major exception, the commentaries all accept our core negative claim, i.e., that autonomy (strictly defined as precarious operational closure) and sense-making (strictly defined in terms of intrinsic normativity) are not, as they are currently formulated and discussed, compatible with massive hetero-scalarity of organizing influ-ences on the activity of living systems.1 This is not to say that they agree with us, but rath-

1 | it is worth being precise here. There are many formulations and interpretations of the con-cept of “autonomy” in and around post-varelian autopoietic enactivism. some of these are less pre-cise, or more open, or less central, or bear less ex-planatory weight than the one we have in mind. our negative argument was that operational clo-sure, under precarious conditions, (a) seems not to obtain as a literal explanation of operational dynamics for some biological systems and (b) can-not be the only source of norms that influence the organization of living systems. This means that modified, extended, or otherwise re-thought con-ceptions of autonomy and intrinsic normativity could avoid this critique; where commenters sug-gest such a course of action, we take them to be

er that this claim precisely – as opposed to a more general antagonism towards enactivist accounts, or other particular issues – is not their target. The second part of the response deals with the one major exception to this, and addresses various related criticisms of our article.

adding to autopoietic enactivism« 3 » We can begin by agreeing with

both Fred Cummins (§§6f) and tom Froese (§§2, 14), who point out that enactive con-cepts are undergoing consistent develop-ment, and that they can be extended and modified to deal with explanatory challeng-es. Further, given the historical progression in enactivism’s research focus from cells to-wards socially and cognitively rich types of human action, it is reasonable that the ex-isting concepts are not (or not yet) perfectly suited to the analysis of human life. indeed, there is no reason our specific critiques of autonomy and sense-making could not be made from “within” enactivism; their pur-pose is to encourage re-consideration of the role played by certain concepts in develop-ing explanations for behaviour.

« 4 » Robert Lowe and nathaniel Barrett both engage in active re-consideration of this kind. Lowe proposes reconceiving en-activism in light of Peter sterling’s (2012) notion of allostasis, described as the key characteristic of mechanisms for the regula-tion of biological activity. importantly, of the three positions he describes (see §2), reactive homeostatic regulation and transient allostat-ic regulation describe systems characterized by operational closure (i.e., a strict concep-

proposing a move away from the use of precarious operational closure as an explanatory concept.



255

explanatory pluralism and precise Conceptual Development matthew Harvey et al.

Enactivism


tion of autonomy), while non-transient al-lostatic regulation describes operationally open systems. on our reading, this means no explanatory weight is given to the notion of operational closure, which is instead used purely as a descriptive tool, where and when that is appropriate. This allays our concerns about the concept’s over-use.

« 5 » Partly for this reason, we are sym-pathetic to Lowe’s position. What he de-scribes as “cybernetic enactivism” goes be-yond post-varela autopoietic enactivism on a decisive point, namely that it takes extrin-sically derived norms into account. For this reason, Lowe suggests that interactivity and enactivism are compatible if the latter is un-derstood in his cybernetic sense, as both ap-proaches acknowledge that an agent makes sense of its world on the basis of both self-derived and other-derived norms. (a satis-fying account of this would amount to cross-ing successfully the so-called cognitive gap, which, as Froese notes, has not been dealt with by autopoietic enactivist approaches.) There is some lack of clarity about how to understand “norms” in this discussion – it is not clear either in Lowe’s paper or in ster-ling’s precisely what they are, whether we ought to speak of them as set or as instanti-ated, and so on – but this is a problem for the future. in general, Lowe’s cybernetic enactiv-ism seems to us a helpful addition to the array of biological perspectives that can be used to analyze phenomena of interest.

« 6 » We are worried about one aspect of Lowe’s paper, however, which is that cy-bernetic enactivism seems to adopt a func-tional perspective on social behavior (see, e.g., §9, and footnotes 3 and 5). Whereas it makes good sense to conceive of some organismic activity in terms of functional descriptions, the same does not hold for hu-man sociality. as one of us has argued else-where (Cowley & Gahrn-andersen 2015), social and cognitive functionalisms are both problematic because they reduce individual sense-making to synchronic coordination (thus leaving aside the vital asymmetry be-tween agent and environment that is central to autopoietic enactivism), and because they leave little room for phenomenology (thus precluding useful examination of lived ex-perience).

« 7 » Barrett proposes a similar answer to our concerns. He sees no reason why

enactivists cannot allow that extrinsic nor-mative influences affect activity without giving up their account of intrinsic nor-mativity (§§10f). additionally, he suggests (§9) that the core of the enactive notion of sense-making is asymmetry of organism-environment coupling rather than the op-erational closure of living systems – that this asymmetry is both more explanatorily im-portant and of greater conceptual interest. Certainly, we have no quarrel with Barrett’s suggestion that a single account might give due attention to both agent-external norma-tive influences and those normative influ-ences arising from processes that constitute the agent. We see our target article as an at-tempt to contribute to just such an account. The trouble is that, as far as we are aware, the concepts of normativity and organism-environment asymmetry are both defined in terms of autonomy, construed strictly as precarious operational closure. according to both Ezequiel di Paolo (2005: 434–437, 443) and Evan Thompson (2007: 145–155), the fundamental asymmetry of living sys-tems with respect to their environments consists precisely in the fact that, through their persistent molecular self-organization and self-production, they distinguish them-selves as coherent entities. in addition, the intrinsic normativity of these systems is identical to their self-production, as ongo-ing self-maintenance becomes a norm that in fact organizes each system’s activity – oth-erwise it dies (for extended discussion see Barandiaran & Egbert 2014; Barandiaran, di Paolo & rohde 2009).

« 8 » For this reason, we have serious doubts about Barrett’s optimistic assertions. organism-environment asymmetry seems neither more nor less important than au-tonomy in this account, because it is identi-cal to it. and neither is it clear that extrinsic norms can be added to an autopoietic enac-tivist account, just because the only way to make sense of “normativity” that is current-ly available in the post-varelian autopoietic framework is in terms of parameter con-straints arising from the self-organizing of autonomous systems.2 other conceptions

2 | an added complication here is that in cas-es where a single organism is constituted of mul-tiple autonomous systems – for instance, when an organism’s nervous system displays self-sustain-

of normativity might be added, such as, for instance, Pierre steiner and John stewart’s contribution that suggests that individuals adapt to social heteronomy through intrin-sic (pre-)dispositions (steiner & stewart 2009: 536f). However, they provide no ex-planation of how an agent comes to form such dispositions in the first place without violating its operational closure. For this reason, we do not see a principled way of ex-plaining human social encounters that keeps the existing conceptions of autonomy and sense-making firmly in place. Barrett asks, in §10, why organism-environment coupling cannot be both self-regulated and heteros-calar. it is crucial to any interactivity-based account that it can be, and is. our concern is that this is very difficult or impossible to capture if “self-regulation” consists of in-trinsic normativity, as that concept seems to preclude heterogeneity of organizing influences. The reason for this is that some heteroscalar activities play out on the basis of extrinsic norms that do not emerge from the agent. so to be clear, our disagreement is with a very specific notion of self-regulation (seen specifically in relation to the paradig-matic aspirations of post-varelian enactiv-ism and to autonomy as a core concept of this particular enactivist approach), and not with self-regulation construed in some other way.

maturana and varela« 9 » This brings us to the direct cri-

tiques of our target article. We begin with those raised by Froese, who clarifies and corrects some of the details of our overview of Humberto Maturana, Francisco varela, and post-varelian autopoietic enactivism. The chief issue here is that our interpreta-tion of the two authors presents Maturana’s position as more open and flexible than that of varela (aligning, respectively, with what we call “organizational closure” and what we call “operational closure”). Froese is concerned that this may be inaccurate, and indeed, that the relationship may be the other way around (see §§4–6, and §§7f, 10f, respectively). We will engage with sev-

ing patterns of neural activity – it is not clear how the norms intrinsic to these various systems can affect one another (Froese, Woodwark & ikegami 2013: 211n5).



Cogn

itiv

e SC

ienC

e Co

nCep

tS in

ena

Ctiv

iSm

256


Enactivism



Enactivism

eral aspects of this issue with a view towards clarifying our target article.

« 10 » We agree that Maturana’s work “lends itself to conflicting interpretations” (§4); perhaps more to the point, it requires active and creative interpretation to derive any clear claims from it at all. This applies especially for his most general ideas, such as the discussion of domains of operations. Froese suggests, and we agree, that the pas-sage he quotes can be read as a clear com-mitment to operational closure in our terms, such that it does make “particular claims about the operational dynamics of living systems”, and does preclude non-molecular processes, and processes outside an organ-ism’s operational bounds, from constituting it as an entity. on this reading, we (in our §17) are far too generous to Maturana in at-tributing to him a view that is compatible with organizational closure – again, in our use of that term.

« 11 » But to be frank, it is difficult to make sense of Maturana’s talk of “non-inter-secting domains” in clear terms. at the very least, it admits of several alternative read-ings. For instance, if read (in the manner Froese suggests) as the claim that processes whose operational constituents include, say, muscles, cannot causally affect processes involving molecules as operational constitu-ents, it seems clearly false (e.g., the beating of an animal’s heart has causal effects on the ongoing health of cells in its extremities; if the heart’s dynamics are disrupted, those cells will quickly find that their molecular milieu lacks sufficient oxygen molecules, and they will die).3 so we read it, instead, as a metaphysical claim concerning the concept of autopoiesis in relation to the experiencing human observer, who is given strict injunc-tion not to think that autopoietic organiza-tion (as a characterization of whole organ-isms) has causal effects on whole-organism

3 | This example is more complicated than it may appear at first. on the one hand, it seems to be internal to the organism’s physiology, and so seems to fit with Froese’s interpretation of Mat-urana. on the other, there is reason to think that heart-rate can become coupled between people (Konvalinka et al. 2011; see discussion in stef-fensen 2015: 11f), suggesting that these muscular contractions may be directly coupled to extra-bodily processes.

events, such as anatomical movement. We read it as a metaphysical claim because it is about the type of existence that living sys-tems have (e.g., “systems as composite enti-ties have a dual existence […],” Maturana 2002: 12), and also because the domains and their relations are defined in terms of distinctions drawn by the experiencing hu-man observer (e.g., “the generative relation that an observer may see between these two domains is a historical relation that the ob-server makes […],” ibid).4

« 12 » in addition to these consider-ations, there is the nature of the idea of auto-poiesis. Maturana writes that “[he] proposed it as an abstraction of [his] biological under-standing aware of the biological knowledge of the times” (ibid: 7). at greater length:

“ My assertion that living systems are molecular autopoietic systems is neither a definition nor an explanatory proposition, it is an abstraction of the operational coherences apparent [i.e., seen by an observer] in the actual living of living systems as molecular systems. Therefore, my assertion that living systems are molecular autopoietic systems (and that there can be autopoietic systems only in the molecular space) is a claim about what consti-tutes living systems, a claim about how they arose, and a claim about how they operate in the prag-matics of their living. Moreover, as i have already said, as i claim that living systems are molecular autopoietic systems i do not make a claim about any particular molecular structure in them, rather i make a claim about the kind of molecular net-work that constitutes them.” (Maturana 2002: 11, our emphases)

« 13 » Given the lack of clarity attendant to his use of “constitutes”, “arose”, and “oper-ate”, the only genuinely clear statement here is that autopoiesis is neither explanatory nor definitional, but is rather an “abstraction” of what is “apparent” to biologists. it is proba-bly worth noting that this arguably reflects a change in perspective from his earlier work. Whereas in these 2002 quotes, Maturana is explicitly defining autopoiesis as an abstrac-tion, in the 1970s and 1980s, he used it as something more of an explanatory concept. For example, he offers an account of nervous

4 | We have chosen quotes from the same pa-per Froese refers to in order to avoid the changes in Maturana’s position over time that he mentions.

system operation in terms of its functional closure, suggesting that it undergoes change on the basis of its self-referential properties (Maturana & varela 1980: 25). in taking au-topoiesis to pick out the universal constitu-tional property that characterizes any living system at all points in time, the later Mat-urana seems to take on a philosophical rath-er than a biological position – and as such, what he says is less restrictive with respect to the actual operations of specific systems in particular situations. in this light, how are we to make sense of this line from the pas-sage Froese quotes?

“ The constitution of living systems as autopoiet-ic systems entails their constitution as organisms as a result of the constitution of their operational boundaries which separate the molecules that dynamically participate in their autopoiesis from those that do not.” (Maturana 2002: 15)

« 14 » if this, too, is to be read in terms of an abstraction away from what biologists observe as they go about studying organ-isms, then

� it does not specify anything about the temporality or direct causal dependen-cies among processes involved in consti-tuting organisms, and

� it must in principle be general enough to account for the wide range of phenom-ena biologists encounter.« 15 » The latter property is conceiv-

ably just a matter of abstraction, so no more needs to be said about it. The former seems to be justified by certain other statements he makes regarding the two domains, or the “dual existence of living systems” (Maturana 2002: 12). The distinction between the do-mains is a distinction between, on the one hand, the abstract autopoietic processes that constitutes the system as living and, on the other hand, the concrete domain wherein the living system works as an organism and interacts with its environment (he calls this the “supra-molecular space”, ibid: 14). He writes, “an organism is an autopoietic sys-tem through its molecular composition, not through its supra-molecular existence” (ibid). it seems reasonable to understand this as the idea that the organism as a whole may in fact be operationally open just as long as extrinsic norms and processes do not bear directly on its (abstract) organiza-




Enactivism

257


Enactivism


tion as a living system. This interpretation fits perfectly well with the doctrine of non-intersecting domains, which is the idea that “the relation between these two domains is not causal; these two domains do not inter-sect, nor do the phenomena which pertain to one occur in the other” (ibid: 12). Con-siderations along these lines led us to say that perhaps our own view might be usefully described as in line with Maturana’s – or at least, with this later, early-2000s Maturana. By contrast, it seems to us that varela – agreeing with, or perhaps following, the earlier Maturana who wrote “The Biology of Cognition” (Maturana & varela 1980) – gives explanatory weight to operational features such as “closure”, whether organi-zational or operational in our terms, and uses them to understand (for example) the function of the nervous system (e.g., varela 1992, 1997).

« 16 » But it is crucially important that we do not get too sidetracked by conflicting readings of these texts. our project was not exegetical in nature. Froese’s response makes it clear that our conception of organizational closure is not necessarily “Maturanian”, and that even the labels we use later – “Matura-na-derived”, “Maturana-inspired” – suggest too close a connection. Let us say this: we take it that living systems are characterized by closure among “an abstract specification of the set of relations that obtain among the system’s repeating component processes so long as the system persists” (§14), such that “circular molecular organization might, without contradiction, be realized opera-tionally by non-circular, constitutively open and environment-involving, massively dis-tributed and multi-scalar processes” (§17). Eating and breathing are processes of this kind. although Maturana would say that they are “in the relational domain”, nonethe-less they are stable, indispensable, enabling conditions for the ongoing molecular self-creation of many lineages.

« 17 » Froese also says that we attribute too little to varela and to post-varelan auto-poietic enactivism, in addition to attributing too much to Maturana. it is worth stating unambiguously that the terms “operational” and “organizational” as we use them in our target article cannot be aligned with “op-erational” and “organizational” closure in the writing of Maturana and varela. Froese

(footnote 1) tells us that the two-word se-quence “organizational closure” was intro-duced by varela (1979), for instance, but we claim that the two together wrote about “op-erational closure” as being a feature of “the organization” of a given organism, or ner-vous system (e.g., Maturana & varela 1980: 25, 26, 88, 128; Maturana & varela 1987: 164); on at least one occasion, Maturana used the phrase “closure in the organiza-tion of the nervous system and the organ-ism” (Maturana 1978: 46). so this seems to us a (merely?) terminological issue. We have found no clear pattern in either author’s use of the two terms, although Froese (personal communication) also suggests that “opera-tional closure” may just indicate that a sys-tem displays recurrent dynamics, whereas “organizational closure” may indicate, in ad-dition, that those dynamics arise due to the system’s self-determined organization.

« 18 » our own distinction – given in §14 and footnote 6 of our target article – is motivated by the usage of david rudrauf and colleagues, who write,

“ Moreover, there are two aspects of closure: orga-nizational, which defines the possible interactions in a ‘static’ circular framework, and operational, i.e., the recurrent dynamics that closure elicits.” (rudrauf et al. 2003: 34, emphases in the original)

« 19 » on their reading, varela’s concept of closure (sometimes called organizational, sometimes operational, according to them) had both of these aspects. our target article proposes a modification to this scheme, on which organization is strictly an abstract characterization of relations among repeat-ing processes (meaning that it does not cap-ture either the dynamics of a system’s activity or non-repeating processes), and operation refers to the actual activity-in-time of a par-ticular specifiable system. Thus systems can be organizationally closed without being operationally closed, using our terminology.

autonomy and normativity« 20 » Froese also argues that the con-

ceptual development by varela, di Paolo, and others that leads to the present notions of autonomy and adaptivity is precisely what make it possible for enactivism to engage with “constitutive dynamics that incorpo-rate normatively organized behaviors that

are distributed and environment-involving” (§8). We agree with his positive evaluation of the emphasis on organism-environment interaction rather than on organisms alone, and we agree that the examples he provides are compelling ones. But our concerns stand. For instance, he refers to the exam-ple given by di Paolo (2009b) of an insect able to survive underwater temporarily by trapping air bubbles against its body; in that paper, di Paolo raises the conceptual possi-bility of the same insect finding a way to re-new the bubbles underwater, and so remain there permanently. such a creature would be “alive and not sensu stricto autopoietic” (ibid: 64), and indeed, not operationally closed, as only the insect’s interactional ac-tivity of collecting oxygen – rather than one of its constitutive processes – would enable its ongoing metabolism. on the basis of this and other examples, we are to believe that enactivism is capable of addressing hetero-geneously constituted multiscalar organiza-tion of activity.

« 21 » But we argue in our target article (§§28f) that non-fictional examples of non-operationally-closed living systems abound. non-autonomous systems are not an excep-tion to the rule, they are utterly ordinary. indeed, it is difficult to conceive of any mul-ti-cellular organism as being strictly opera-tionally closed, and even harder to imagine how we might ascertain whether this is the case. and this is without taking into account our concern that many or most organisms are constitutively multiscalar, such that their behaviour takes into account norms and values extrinsic to the agent itself. on our view, non-autonomy ought to be the default assumption for researchers interested in the sorts of phenomena that enactivism, and interactivity-based approaches, are inter-ested in.

« 22 » We feel that Froese’s other exam-ples support this view. For instance, work on participatory sense-making (beginning with de Jaegher & di Paolo 2007) has focused on emergent self-organization in the dynamics of face-to-face interaction. People’s move-ments in these cases can evoke movements in their partners, leading to partially self-sustaining patterns of activity, such that the interaction itself can be seen as an autono-mous system, with its own intrinsic norms (at least, the norm of self-persistence). For



Cogn

itiv

e SC

ienC

e Co

nCep

tS in

ena

Ctiv

iSm

258


Enactivism



Enactivism

Froese, varela’s substrate-neutral concep-tion of autonomy is very helpful here, as “it permits us to study individual behaviors and social interaction dynamics in terms of two interdependent yet mutually irreducible domains” (§11). But are the “social interac-tion dynamics” actually a different “domain” from the movements of the interaction par-ticipants? (again, “domain” is an unclear term; “different” here means, in Froese’s words, “irreducible” yet “interdependent”). do the movements exist as non-social activ-ity? More importantly, when human move-ment does not form a set of hierarchically arranged autonomous systems – say, when someone goes for a run – can it still be social behavior? di Paolo (2009b: 54) and Froese (§10) both propose, similarly, that habits can be seen as self-sustaining, circularly organ-ized patterns of activities. This may be cor-rect, but what advantage is gained by noting it? it seems to us that the generalized no-tion of autonomy is little or no explanatory work in these cases. it serves only to direct researchers to areas of behaviour that seem to comprise circularly organized systems, as opposed to behaviour that is more variously organized.

« 23 » a related concern is voiced by Barrett, who says several times (most sig-nificantly in §9) that he sees no reason to read definitions of autonomy, adaptivity, and sense-making as being monoscalar con-cepts. He is right in the sense that, as far as we are aware, they have only once5 been explicitly discussed in relation to the ques-

5 | di Paolo (2005: 444f) offers a charac-teristically clear and rigorous account of the “minimum temporal granularity” established by a system’s ability to regulate its activity as part of self-maintenance. That is, when an organism’s ac-tivity operates to correct a tendency that threatens its autonomy, the activity takes on the structure of an action: it is initiated for a specifiable reason, it succeeds or fails in alleviating the conditions that triggered it, and at some point the organism’s ac-tivity transitions such that it is no longer regulated according to those conditions. our two worries about this account will be predictable: (a) it defi-nitionally links the temporality of living processes to the conservation of autonomy, and (b) it only mentions the temporality defined by the initia-tion, persistence, and cessation of directed actions an autonomous individual undertakes.

tion of timescales. However, on the one hand, they are exemplified using examples of monoscalar coupling, e.g., the coupling of bodily movements (de Jaegher & di Paolo 2007), the coupling of bacterial flagellation to sucrose gradients by means of chemo-sensation (di Paolo 2005: 445; Thompson 2007: 147), and so on. second, coupling in general is a matter of constraints on the pa-rameters of one system by the activity of an-other; the timescales on which it takes place are therefore determined by the dynamics of the coupled systems. and again, enactive discussions of autonomy, adaptivity, and so forth have focused on cellular regulation of ion concentrations, nervous system activity, and anatomical movement, usually one at a time. This means that, just as a matter of fact, they have tended to focus on coupling on single timescales. it is possible to define operational closure such that it applies to processes on multiple timescales, but it is not clear whether the relevant properties, such as “self-production”, still make sense in these cases.

« 24 » There is also another side to this issue, which is to treat it as a question about the “here and now” rather than about mon-oscalarity. according to Barrett, we do not provide any evidence that autopoietic en-activism confines itself to a situated, “in the moment” perspective. However, it is at least worth recognizing that in the enactive litera-ture, there are many examples of focus and emphasis lying on the here and now, even if there are not explicit restrictions on what is to be considered as an explanatory resource. varela argued explicitly in favour of a situat-ed enactive approach, placing strong empha-sis on the emergence of phenomenological experience from moment to moment (see, for instance, varela & shear 1999: 115f).6 He underlines his position in other writ-ings (e.g., varela 1987; varela, Thompson & rosch 1991) using the metaphor of “lay-ing down a path in walking”. More recently, other enactivists including Hanne de Jaegh-er and di Paolo (2007) and Elena Cuffari (2014) have argued extensively that human social life plays out on the immediate level of inter-bodily coordinations and/or through

6 | as Froese notes (personal communica-tion), this is another legacy of Maturana’s early writings (e.g., Maturana & varela 1980: 18).

in situ linguistic interactions. With respect to language, Cuffari writes the following:

“ although we are trained, as scholars and us-ers of language, to attend and deal primarily with words, i suggest that deeper listening and more mindful engagement result when we understand wordings as in-the-moment interactions with ex-perience.” (Cuffari 2014: 211f)

« 25 » However, linguistic activity, or languaging, is also constrained by factors that do not play out in a situation but con-dition it nonetheless. This limited focus of autopoietic enactivism means that topics re-lated to languaging are excellent subjects for interactivity-based approaches.

Questions of metaphysics« 26 » There is one other issue we would

like to address, which is Cummins’s challenge to the “objective key” of our approach to ex-planation, and to our metaphysical stance in general. in particular, he is concerned that we ignore that enaction is not “part of an objective science of biology, concerned with individuated discrete organisms” (§2). to further his argument, he contrasts “the man-ner in which creation is treated in Buddhist and in Christian cosmologies” (§3), with a starting point in the Buddhist “core concept of dependent origination (pratityasamutpa-da)” (§4), and the two views on time related to John Ellis Mctaggart’s (1908) distinction between linear time and the lived experi-ence of time. Cummins’s main point seems to be that “[a]ppeals to ‘lived experience’ will […] run into opposition if they feature in scientific discourse conducted in an objec-tive key” (§3), because, he argues, “[i]n an objective key, the world comes first, and experience is a puzzle to be understood as something within the world (though never found)”. By contrast, “[i]n a strongly con-structivist, enactive key, experience is prior to any distinction between subject and world” (§4).7 Cummins is to be applauded for addressing the big issues in this debate, and in particular for highlighting the Bud-dhist philosophical thinking in varela et al. (1991). However, we remain unconvinced by his argument.

7 | The expression “prior to” is odd, since that actually presupposes a linear concept of time.




Enactivism

259


Enactivism


« 27 » to begin with, Cummins misrep-resents our position in the following state-ment:

“ […] HGs [i.e., we] speak of an organism’s ex-perience as ‘being coupled to the world’ (§13, em-phasis mine [i.e., Cummins’s]). if subject and world co-arise, then the ‘world’ of which we speak is not […] something that exists in insolent opposition to the subject.” (§5)

notice that we did not talk about “an organ-ism’s experience as ‘being coupled to the world’” (our emphasis) but about “organ-isms’ experience of being coupled to the world” (§13, our emphasis). Clearly, there is nothing contradictory in a model of de-pendent origination and a model of experi-ence where one part of the co-arisen dyad experiences its coupling to the other part. in fact, were that not the case, the Buddha’s doctrine would have been empty specula-tion; its explanatory power comes from the fact that it builds on his analysis of his expe-rience as something that is not “his” (in the meaning of something private, apart from the world) but a relational and fluctuating aspect of the co-arisen dyad. Borrowing the words of Eleanor rosch, we must un-derstand experience in terms of “a seamless whole or seamless web, in which perceiver/categorizer and perceived/categorized are simply opposite poles of the same event” (rosch 1999: 71).

« 28 » overall, we find that Cummins misreads our commitment to (some vari-ety of) naturalism and realism. Cummins relies on a rhetorical model that contrasts Buddhism and enactivism with the “usual suspects”: the “Copernican-Galilean-new-tonian-Cartesian world view” (§3), which “has no place for mind” (§3), and which takes Christianity as its “unacknowledged [metaphysical] background” (§3). Between these two positions, tertium non datur. However, that is a non sequitur. our interac-tivity-based approach is exactly an attempt to reconcile a naturalised world view with an emphasis on life and experience. in short, the interactivity-based approach takes a starting point in an extended ecology (stef-fensen 2011) that represents neither a mind-less, mechanistic ontology nor the view that experience occupies a privileged position in our ecologically embedded existence.

« 29 » as presented in the target ar-ticle, our approach presupposes the no-tion of temporal multiscalarity. This notion contrasts with Cummins’ claim that “subject and world arise in the always-present now” (§4). to understand fully what is at stake here, let us follow Cummins’ route and con-sider some of the Buddhist philosophical underpinnings. if we accept the doctrine of paticcasamuppada,8 it follows that all co-arisen phenomena are in a state of im-permanence (anicca), because “things” that are co-arisen do not owe their existence to any inner, transcendent qualities – they are suññatā (empty), as described in the Bud-dhist doctrine of anatta (no-self). The key to understanding this philosophical system is to appreciate that it is primarily an attempt to grasp the temporal dynamics that play out in our lives. Paticcasamuppada, on this view, is a theory of causality, rather than a cosmol-ogy. it argues that our being is an interbeing and that the temporal flow of our existence is not tied to autonomous bodies, because they, in themselves, are “empty.”

« 30 » on this view, no causal power stands outside what one of us has termed “the paticca field,” i.e., the open-ended sys-tem of co-arisen entities in the world.9 in later Buddhist philosophy, most notably in the Chinese Huayan tradition (Fox 2009), the term interpenetration has been used to illustrate this radical view of interbeing (cf. the metaphor of Indra’s net). Giving expe-rience a privileged metaphysical status, as Cummins does, thus seems misguided. Ex-perience is no less and no more than an explanans: it too must be grasped within a larger network of existence. it is for this rea-son that we adhere to a naturalist position: to understand experience requires that we understand the causal interplay, or the in-terpenetration, between experience and the Paticca field at large. However, since some of the temporal dynamics in the Paticca field happen infinitely faster, and others infinitely slower, than we can experience, we see no

8 | Following david Kalupahana (1992), we use the Pali terms for Buddhist concepts.

9 | Cf. the presentation “Event analysis in a distributed health interaction” by one of us [s. v. s.] at the conference “Grounding language in perception and (inter)action,” Gordon College, Wenham Ma, June 2009.

way in which a first-person perspective can grasp that interplay and interpenetration. Borrowing the words from the norwegian poet Bertrand Besigye: “og du dør så lang-somt at du tror du lever” (“and you die so slowly that you believe you live”). in fact, the pivot of the Buddha’s philosophy is to show that pain and suffering (dukkha) arise be-cause experience and reality are not tightly coupled or synchronized.

« 31 » The Paticca field has no spa-tiotemporal limits, and thus no center. it is multi-logical, multi-scalar, and multi-dimensional, and it gives rise to a flux of transient, interrelated, and interdependent events that occur in series of arising, being and ceasing (“the aBC of impermanence”). one such event is the co-arising of experi-ence. We trace experience to the emergence of life, i.e., a set of irreversible temporal dy-namics that de-synchronises the interbeing of organism and environment (experiencer and world). Thus, understanding the exact temporal dynamics that play out between different complex systems, and hence be-tween organisms and their environment(s), is a way of establishing a multiscalar frame-work for understanding life. These dynam-ics may vary in complexity, and human cognition (and human lived experience) is vastly complex, and not just because it plays out on more timescales than, say, those of Chlamydomonas nivalis. What is so aston-ishing about human life is that human be-ings have learnt not only to enact different timescales but to “create and inhabit their own temporal domains” (Cowley & stef-fensen 2015: 490), which has made us into time-rangers (ibid), agents capable of flexibly operating in many different temporal scales.

« 32 » With respect to Cummins’s com-ments, then, we hope to have made clear that we are not indiscriminately blending two incompatible ontologies. While the above account is obviously sketchy and super-ficial, we hope that it indicates our overall approach to understanding how “lived ex-perience” can, and indeed should, be traced to the totality, the Paticca field, from which experiencer and world are co-arisen, and not to the “world” as an objective phenom-enon outside the experiencer. to the degree that such a model can be further developed, we maintain that our interactivity-based ap-proach has potential for understanding lived



Cogn

itiv

e SC

ienC

e Co

nCep

tS in

ena

Ctiv

iSm

260

Combined References matthew Harvey et al.


experience against the backdrop of a natu-ralised and realist ontology.

« 33 » There is a great deal in the com-mentaries that we have left unanswered, in the interests of keeping this response shorter than the original target article. The upshot, we think, is very positive. We see no reason to adhere strictly – or to reject – post-varelan autopoietic enactivism, or its core concepts. Better by far to accept the co-incidence of various alternative models, which provide complementary means of ac-counting for wide ranges of organizational influences. in effect, we suggest making the move recommended by Barrett (to stipu-late that external norms, as well as intrinsic

ones, play a role in sense-making), which might be done using an interactivity-based framework, an allostasis based framework as proposed by Lowe, or a different frame-work altogether, in an attempt to develop a clearer picture of how groups of people collaborate in collectively organizing activ-ity. to borrow Froese’s way of putting it, we are after “an explanation of how, in that very moment, the operations underlying the in-dividual’s choice of action were realized in accordance with […] cultural norms” (§14). it is still not clear to us whether this can be achieved within the post-varelian enactive paradigm. We still have strong doubts that it can, so long as autonomy and sense-making

remain, unchanged, at the center of that paradigm’s conceptual apparatus. But as we have discussed, the concepts are flexible; the method that has been used to develop them can be used to develop them further, and po-tentially to apply them to new phenomena in light of our criticisms. and in addition to this, it is clear that other enactivist concepts, approaches, and perspectives are helpful in their own right, and will contribute to our understanding of cognitive activity.

received: 29 February 2016 accepted: 6 March 2016

Combined References

Ashby W. R. (1952) design for a brain. Chapman and Hall, London.

Attanasi A., Cavagna A., Del Castello L., Giardina I., Grigera T. S., Jelić, A., Melillo S., Parisi L., Pohl O., Shen E. & Viale M. (2014) information transfer and behavioural inertia in starling flocks. nature Physics 10: 691–696.

Baber C. (2015) Thinking through tools: What can tool-use tell us about distributed cognition? studies in Logic, Grammar and rhetoric 41(1): 25–40.

Baber C., Parekh M. & Cengiz T. G. (2014) tool use as distributed cognition: How tools help, hinder and define manual skill. Frontiers in Psychology 5: 116. http://journal.frontiersin.org/article/10.3389/fpsyg.2014.00116/

Bailly F. & Longo G. (2008) Extended critical situations: The physical singularity of life phenomena. Journal of Biological systems 16(2): 309–336.

Barandiaran X. E. (2012) Pushing the ac-celerator on enactive perception: How senso-rimotor dynamics can constitute minds. seminario interuniversitario de Ciencia Cognitiva, XXii.

Barandiaran X. E. & Di Paolo E. A. (2014) a ge-neological map of the concept of habit. Fron-tiers in Human neuroscience 8: 522. http://journal.frontiersin.org/article/10.3389/fnhum.2014.00522

Barandiaran X. E., Di Paolo E. A. & Rohde M. (2009) defining agency: individuality, nor-mativity, asymmetry and spatio-temporality in action. Journal of adaptive Behavior 17(5): 367–386. ▶︎ http://cepa.info/324

Barandiaran X. E. & Egbert M. D. (2014) norm-establishing and norm-following in autono-mous agency. artificial Life 20: 5–28.

Barrett N. F. (2015a) Enactive theory and the problem of non-sense. adaptive Behavior 23: 234–240. ▶︎ http://cepa.info/2472

Barrett N. F. (2015b) The normative turn in enactive theory: an examination of its roots and implications. topoi. online first. ▶︎ http://cepa.info/2473

Barrows W. M. (1915) The reactions of an orb-weaving spider, Epeira sclopetaria Clerck, to rhythmic vibrations of its web. Biological Bulletin 29(5): 316–326+328.

Barth F. G. (1985) neuroethology of the spider vibration sense. in: F. G. Barth (ed.) neu-robiology of arachnids. springer-verlag, Berlin: 203–229.

Bingham G. P. (1988) task-specific devices and the perceptual bottleneck. Human Move-ment science 7: 225–264.

Buhrmann T. & Di Paolo E. (2014) non-representational sensorimotor knowledge. in: del Pobil a. P, Chinellato E., Ester Martinez-Martin E., Hallam J., Cervera E. & Morales a. (eds.) From animals to animats 13. springer, new York: 21–31. ▶︎ http://cepa.info/2521

Buhrmann T., Di Paolo E. & Barandiaran X. (2013) a dynamical systems account of sensorimotor contingencies. Frontiers in Psychology 4: 285. ▶︎ http://cepa.info/2386

Cannon W. B. (1929) organization for physi-ological homeostasis. Physiological reviews 9(3): 399–431.

Cavagna A., Cimarelli A., Giardina I., Parisi G., Santagati R. & Stefanini F., Massimiliano. (2010) scale-free correlations in starling flocks. Pnas 107(26): 11865–11870.

Colombetti G. (2014) The feeling body. Mit Press, Cambridge Ma. ▶︎ http://cepa.info/777

Cowley S. J. (2009) distributed language and dynamics. Pragmatics and Cognition 17(3): 495–507.

Cowley S. J. (2011) distributed language. John Benjamins, Philadelphia Pa.

Cowley S. J. (2016) Entrenchment. in: schmid H. J. (ed.) Entrenchment, memory and automaticity: The psychology of linguis-tic knowledge and language learning. de Gruyter, new York. in: Press.

Cowley S. J. & Gahrn-Andersen R. (2015) deflating autonomy: Human interactivity in the emerging social world. intellectica 63: 49–63.

Cowley S. J & Nash L. (2013) Language, inter-activity and solution probing: repetition without repetition. adaptive Behavior 21(3): 187–198.



http://journal.frontiersin.org/article/10.3389/fpsyg.2014.00116/

http://journal.frontiersin.org/article/10.3389/fpsyg.2014.00116/

http://journal.frontiersin.org/article/10.3389/fnhum.2014.00522



http://cepa.info/324






261


Enactivism


Cowley S. J. & Steffensen S. V. (2015) Coordina-tion and language: temporality and time-ranging. interaction studies 16(3): 474–494.

Cowley S. J. & Vallée-Tourangeau F. (eds.) (2013) Cognition beyond the brain. springer, London.

Craig A. D. (2015) How do you feel? an intero-ceptive moment with your neurobiological self. Princeton university Press, Princeton.

Cuffari E. C. (2014) on being mindful about misunderstandings in languaging: Making sense of non-sense as the way to sharing linguistic meaning. in: Cappuccio M. & Froese t. (eds.) Enactive cognition at the edge of sense-making: Making sense of non-sense. Palgrave Macmillan, Basingstoke uK: 207–237.

Cuffari E. C., Di Paolo E. & De Jaegher H. (2014) From participatory sense-making to language: There and back again. Phenom-enology and the Cognitive sciences 14(4): 1089–1125. ▶︎ http://cepa.info/2516

Cummins F. (2009) rhythm as an affordance for the entrainment of movement. Phonetics 66: 15–28.

Cummins J. (2012) do the origins of biologi-cal general intelligence lie in an adaptation of the stress response? in: simeonov P. L., smith L. s.& andrée C. Ehresmann a. C. (eds.) integral biomathics. springer, Berlin: 155–168.

Cummins F. (2013) towards an enactive account of action: speaking and joint speaking as exemplary domains. adaptive Behavior 21(3): 178–186.

De Jaegher H. & Di Paolo E. A. (2007) Participa-tory sense-making: an enactive approach to social cognition. Phenomenology and the Cognitive sciences 6(4): 485–507. ▶︎ http://cepa.info/2387

De Jaegher H., Di Paolo E. & Gallagher S. (2010) Can social interaction constitute so-cial cognition? trends in Cognitive sciences 14(10): 441–447.

De Jaegher H. & Froese T. (2009) on the role of social interaction in individual agency. adaptive Behavior 17: 444–460.

De Jesus: (2015) autopoietic enactivism, phenomenology and the deep continu-ity between life and mind. Phenomenol-ogy and the Cognitive sciences, 1–25. ▶︎ http://cepa.info/2385

Degenaar J. & Myin E. (2014) representation-hunger reconsidered. synthese 191: 3639–3648.

Di Paolo E. A. (2003) organismically-inspired robotics: Homeostatic adaptation and teleol-ogy beyond the closed sensorimotor loop. in: Murase K. & asakura t. (eds.) dynami-cal systems approach to embodiment and sociality. advanced Knowledge international, adelaide: 19–42. ▶︎ http://cepa.info/2514

Di Paolo E. A. (2005) autopoiesis, adaptiv-ity, teleology, agency. Phenomenology and the Cognitive sciences 4(4): 429–452. ▶︎ http://cepa.info/2269

Di Paolo E. A. (2009a) Extended life. topoi 28(1): 9–21. ▶︎ http://cepa.info/322

Di Paolo E. A. (2009b) overcoming autopoiesis: an enactive detour on the way from life to society. in: Magalhães r. & sanchez r. (eds.) autopoiesis in organization theory and practice. Emerald, Bingley uK: 43–68. ▶︎ http://cepa.info/2366

Di Paolo E. A. (2015) interactive time-travel: on the intersubjective retro-modulation of intentions. Journal of Consciousness studies 22(1–2): 49–74.

Di Paolo E. A., Rohde M. & De Jaegher H. (2010) Horizons for the enactive mind. in: stewart J., Gapenne o. & di Paolo E. (eds.) Enaction: toward a new paradigm for cogni-tive science. Mit Press, Cambridge Ma: 33–88. ▶︎ http://cepa.info/2413

Dretske F. (1988) Explaining behavior. Mit Press, Cambridge Ma.

Ellis N. C. & Larsen-Freeman D. (2006) Lan-guage emergence: implications for applied linguistics – introduction to the special issue. applied Linguistics 27(4): 558–589.

Enfield N. J. (2011) sources of asymmetry in hu-man interaction. in: stivers t., Mondada L. & steensig J. (eds.) The morality of knowledge in conversation. Cambridge university Press, Cambridge: 285–312.

Enfield N. J. (2013) relationship thinking: agency, enchrony, and human sociality. oxford university Press, oxford.

Enfield N. J. (2014) Causal dynamics of language. in: Enfield n. J., Kockelman P. & sidnell J. (eds.) The Cambridge handbook of linguistic anthropology. Cambridge university Press, Cambridge: 325–342.

Fischer A. H. & Mansted A. (2008) social func-tions of emotion. in: Lewis M., Haviland-Jones J. & Barrett L. (eds.) Handbook of emo-tions. The Guilford Press, London: 456–470.

Fowler C. A. (2014) talking as doing: Language forms a public language. new ideas in Psy-chology 32: 174–182.

Fox A. (2009) dushun’s huayan fajie guan men (Meditative approaches to the huayan dharmadhatu). in: Edelglass W. & Garfield J. L. (eds.) Buddhist philosophy: Essential readings. oxford university Press, oxford: 73–82.

Froese T. & Di Paolo E. A. (2009) sociality and the life-mind continuity thesis. Phenom-enology and the Cognitive sciences 8(4): 439–463.

Froese T. & Di Paolo E. A. (2010) Modelling social interaction as perceptual crossing: an investigation into the dynamics of the interaction process. Connection science 22(1): 43–68.

Froese T. & Di Paolo E. A. (2011) The enactive approach: Theoretical sketches from cell to society. Pragmatics and Cognition 19(1): 1–36. ▶︎ http://cepa.info/2367

Froese T. & Fuchs T. (2012) The extended body: a case study in the neurophenomenol-ogy of social interaction. Phenomenology and the Cognitive sciences 11: 205–235. ▶︎ http://cepa.info/2389

Froese T. & Gallagher S. (2012) Getting inter-action theory (it) together: integrating developmental, phenomenological, enactive, and dynamical approaches to social interac-tion. interaction studies 13(3): 436–468. ▶︎ http://cepa.info/2368

Froese T. & Stewart J. (2010) Life after ashby: ultrastability and the autopoietic founda-tions of biological autonomy. Cybernet-ics & Human Knowing 17(4): 7–49. ▶︎ http://cepa.info/387

Froese T. & Stewart J. (2012) Enactive cognitive science and the biology of cognition: a response to Humberto Maturana. Cyber-netics and Human Knowing 19(4): 61–74. ▶︎ http://cepa.info/2388

Froese T., Virgo N. & Ikegami T. (2014) Motility at the origin of life: its characterization and a model. artificial Life 20(1): 55–76.

Froese T., Woodward A. & Ikegami T. (2013) turing instabilities in biology, culture, and consciousness? on the enactive origins of symbolic material culture. adaptive Behav-ior 21(3): 199–214.

Froese T. & Ziemke T. (2009) Enactive artificial intelligence. artificial intelligence 173(3): 466–500. ▶︎ http://cepa.info/279

Fuchs T. (2013) The phenomenology and development of social perspectives. Phe-nomenology and the Cognitive sciences 12: 655–683.

















Cogn

itiv

e SC

ienC

e Co

nCep

tS in

ena

Ctiv

iSm

262


Gallagher S. (2013) The socially extended mind. Cognitive systems research 25–26: 4–12. ▶︎ http://cepa.info/2483

Gallagher S. & Hutto D. (2008) understand-ing others through primary interaction and narrative practice. in: Zlatev J., racine t. P., sinha C. & itkonen E. (eds.) The shared mind: Perspectives on intersubjectivity. John Benjamins, amsterdam: 17–38.

Gangopadhyay N. & Kiverstein J. (2008) Enactivism and the unity of perception and action. topoi 28: 63–73.

Goldstein L. & Fowler C. A. (2003) articulatory phonology. in: schiller n. o. & Meyer a. s. (eds.) Phonetics and phonology in language comprehension and production: differences and similarities. Mouton de Gruyter, Berlin: 159–207.

Gordon D. M. (2010) ant encounters. Princeton university Press, Princeton nJ.

Griffiths E. & Scarantino A. (2005) Emotions in the wild. in: robbins P. & aydede M. (eds.) The Cambridge handbook of situated cognition. Cambridge university Press, Cambridge: 437–453.

Gu X. & FitzGerald T. H. (2014) interocep-tive inference: Homeostasis and decision-making. trends in Cognitive sciences 18(6): 269–270.

Harvey M. (2015) Content in languaging: Why radical enactivism is incompatible with rep-resentational theories of language. Language sciences 48: 90–129.

Havelange V. (2010) The ontological constitu-tion of cognition and the epistemological constitution of cognitive science: Phe-nomenology, enaction, and technology. in: stewart J., Gapenne o. & di Paolo E. a. (eds.) Enaction: toward a new paradigm for cognitive science. Mit Press, Cambridge Ma: 335–359. ▶︎ http://cepa.info/2466

Heras-Escribano M., Noble J. & De Pinedo M. (2014) Enactivism, action and normativity: a Wittgensteinian analysis. adaptive Behav-ior 23(1): 20–33.

Hoppitt W. & Laland K. N. (2013) social learn-ing. Princeton university Press, new York.

Hurley S. (2001) Perception and action: alterna-tive views. synthese 129(1): 3–40.

Hurley S. (2002) Concsciousness in action. Har-vard university Press, Cambridge Ma.

Hutchins E. (2014) The cultural ecosystem of human cognition. Philosophical Psychology 27(1): 34–49.

Hutto D. D. (2005) Knowing what? radical ver-sus conservative enactivism. Phenomenol-ogy and the Cognitive sciences 4: 389–405. ▶︎ http://cepa.info/2391

Hutto D. D. (2013) Exorcising action oriented representations: ridding cognitive science of its nazgûl. adaptive Behavior 21(3): 142–150.

Hutto D. D. (2015) overly enactive imagina-tion? radically re-imagining imagining. The southern Journal of Philosophy 53(s1): 68–89.

Hutto D. D., Kirchhoff M. & Myin E. (2014) Extensive enactivism: Why keep it all in? Frontiers in Human neuroscience 8: 706. ▶︎ http://cepa.info/2392

Hutto D. D. & Myin E. (2013) radicalizing enac-tivism. Mit Press, Cambridge Ma.

Hutto D. D. & Satne G. (2015) The natural ori-gins of content. Philosophia 43(3): 521–536.

Iizuka H., Ando H. & Maeda T. (2013) Extended homeostatic adaptation model with meta-bolic causation in plasticity mechanism: toward constructing a dynamic neural network model for mental imagery. adaptive Behavior 21(4): 263–273.

Jensen T. W. (2014) Emotion in languaging: Language and emotion as affective, adaptive and flexible behavior in social interaction. Frontiers in Psychology 5: 720.

Johnson G. J., Bruner II G. C. & Kumar A. (2006) interactivity and its facets revisited: Theory and empirical test. Journal of adver-tising 35(4): 35–52.

Kalupahana D. (1992) a history of Buddhist philosophy. Continuities and discontinuities. university of Hawaii Press, Honolulu.

Kello C. T. & Van Orden G. C. (2009) soft-as-sembly of sensorimotor function. nonlinear dynamics, Psychology, and Life sciences 13(1): 57–73.

Kelso J. A. S. (1995) dynamic patterns: The self-organization of brain and behaviour. Mit Press, Cambridge.

Kirsh D. (1997) initeractivity and multimedia interfaces. instructional science 25: 79–96.

Kirsh D. (2015) The importance of chance and interactivity in creativity. Pragmatics and Cognition 22(1): 5–26.

Kiverstein J. & Rietveld E. (2015) The primacy of skilled intentionality: on Hutto and satne’s The natural origins of Content. Philosophia 43(3): 1–21. http://link.springer.com/article/10.1007/s11406–015–9645-z/fulltext.html

Konvalinka I., Xygalatas D., Bulbulia J., Schjødt U., Jegindø E.-M. Wallot S. & Roepstorff A. (2011) synchronized arousal between performers and related specta-tors in a fire-walking ritual. Pnas 108(20): 8514–8519.

Laland K. N., Odling-Smee J. & Feldman M. W. (2000) niche construction, biological evolution, and cultural change. Behavioral and Brain sciences 23: 131–175.

Latour B. (2013) an inquiry into modes of existence. translated by Catherine Porter. Harvard university Press, Cambridge Ma. French original published in 2012.

Lemke J. (2000) across the scales of time: arti-facts, activities, and meanings in ecosocial systems. Mind, Culture, and activity 7(4): 273–290.

Loughlin V. (2014) sensorimotor knowledge and the radical alternative. in: Bishop J. M. & Martin a. o. (eds.) Contemporary sensorimotor theory. springer, Heidelberg: 105–116.

MacWhinney B. (2005) The emergence of linguistic form in time. Connection science 17(3–4) 191–211.

Malafouris L. (2013) How things shape the mind. Mit Press, Cambridge Ma.

Malafouris L. (2015) Metaplasticity and the primacy of material engagement. time and Mind 8(4): 351–371.

Maturana H. R. (1970) Biology of cognition. Biological Computer Laboratory (BCL) research report BCL 9.0. university of il-linois, urbana. ▶︎ http://cepa.info/535

Maturana H. R. (1978) Biology of language: The epistemology of reality. in: G. a. Miller & E. Lenneberg (eds.) Psychology and biology of language and thought: Essays in honor of Eric Lenneberg. academic Press, new York: 27–63. ▶︎ http://cepa.info/549

Maturana H. R. (1988) ontology of observ-ing: The biological foundations of self-consciousness and the physical domain of existence. in: donaldson r. E. (ed.) texts in cybernetic theory: an in-depth exploration of the thought of Humberto Maturana, Wil-liam t. Powers, and Ernst von Glasersfeld. american society for Cybernetics, Felton Ca. ▶︎ http://cepa.info/597

Maturana H. R. (2002) autopoiesis, structural coupling, and cognition: a history of these and other notions in the biology of cogni-tion. Cybernetics and Human Knowing 9(3–4): 5–34. ▶︎ http://cepa.info/685







http://link.springer.com/article/10.1007/s11406






263


Enactivism


Maturana H. R. (2011) ultrastability… auto-poiesis? reflective response to tom Froese and John stewart. Cybernetics and Human Knowing 18(1–2): 143–152.

Maturana H. R. & Poerksen B. (2004) From being to doing. Carl-auer verlag. German original published in 2001.

Maturana H. R. & Varela F. J. (1980) autopoiesis and cognition. reidel, dordecht.

Maturana H. R. & Varela F. J. (1987) The tree of knowledge. shambhala Publications, Boston Ma.

McMillan S. J. (2006) Exploring models of in-teractivity from multiple research traditions. in: Lievrouw L. a. & Livingstone s. (eds.) The Handbook of new Media. sage, London: 205–230.

McTaggart J. M. E. (1908) The unreality of time. Mind 17: 456–473. http://www.ditext.com/mctaggart/time.html

Menary R. (2010) introduction to the special issue on 4E cognition. Phenomenology and the Cognitive sciences 9: 459–463. ▶︎ http://cepa.info/2284

Merker B. & Okanoya K. (2007) The natural history of human language. in: Lyon C., ne-haniv C. L. & Cangelosi a. (eds.) Emergence of Communication and Language. springer London: 403–420.

Millikan R. (2004) varieties of Meaning. Mit Press, Cambridge Ma.

Miłkowski M. (2015a) satisfaction conditions in anticipatory mechanisms. Biology & Phi-losophy 30(5): 709–728. http://link.springer.com/article/10.1007/s10539–015–9481–3/fulltext.html

Miłkowski M. (2015b) The hard problem of content: solved (long ago). studies in Logic, Grammar and rhetoric 41(1): 73–88.

Mossio M. & Taraborelli D. (2008) action-dependent perceptual invariants: From ecological to sensorimotor approaches. Con-sciousness and Cognition 17: 1324–1340.

Muntean I. & Wright C. D. (2007) autonomous agency ai, and allostasis: a biomimetic perspective. Pragmatics & Cognition 15(3): 485–513.

Myin E. & Hutto D. (2015) rEC: Just radical enough. studies in Logic, Grammar and rhetoric 41(1): 61–71.

Myin E. & O’Regan K. (2002) Perceptual consciousness, access to modality and skill theories. a way to naturalize phenomenol-ogy? Journal of Consciousness studies 9(1): 27–46.

Noë A. (2004) action in perception. Mit Press, Cambridge Ma.

Noë A. (2009) out of our heads. Macmillan, new York.

Noë A. (2012) varieties of presence. Harvard university Press, Cambridge Ma.

O’Regan K. (2011) Why red doesn’t sound like a bell. oxford university Press, oxford.

O’Regan K., Myin E. & Noë A. (2005) sensory consciousness explained (better) in terms of “corporality”and “alerting capacity.” Phenomenology and the Cognitive sciences 4(4): 369–387.

O’Regan K. & Noë A. (2001a) a sensorimotor account of vision and visual consciousness. Behavioral and Brain sciences 24: 939–1031. ▶︎ http://cepa.info/2285

O’Regan K. & Noë A. (2001b) What is it like to see: a sensorimotor theory of percep-tual experience. synthese 129(1): 79–103. ▶︎ http://cepa.info/2394

Pedersen S. B. (2012) interactivity in health care: Bodies, values and dynamics. Language sci-ences 34(5): 532–542.

Port R. F. (2010a) Language as a social institu-tion: Why phonemes and words do not live in the brain. Ecological Psychology 22: 304–326.

Port R. F. (2010b) rich memory and distributed phonology. Language sciences 32: 43–55.

Prinz J. (2006) Putting the brakes on enac-tive perception. Psyche 12(1): 1–19. ▶︎ http://cepa.info/2395

Rączaszek-Leonardi J. (2014) Multiple systems and multiple timescales of language dynam-ics: Coping with complexity. Cybernetics and Human Knowing 21(1–2): 37–52.

Rączaszek-Leonardi J. & Cowley S. (2012) The evolution of language as controlled collectiv-ity. interaction studies 13(1): 1–16.

Ramscar M. & Port R. F. (2016) How spoken languages work in the absence of an inven-tory of discrete units. Language sciences 53: 58–74.

Ratcliffe M. (2012) Feelings of being. oxford university Press, oxford.

Rosch E. (1999) reclaiming concepts. in: núñez r. & Freeman W. J. (eds.) reclaiming cogni-tion. imprint academic, Thorverton uK: 61–78.

Rowlands M. (2010) The new sciences of the mind. Mit Press, Cambridge Ma.

Rudrauf D., Lutz A., Cosmelli D., Lechaux J.-P. & Le van Quyen M. (2003) From autopoi-esis to neurophenomenology: Francisco

varela’s exploration of the biophysics of being. Biological research 36: 27–65. ▶︎ http://cepa.info/1140

Ruiz-Mirazo K. & Moreno A. (2012) au-tonomy in evolution: From minimal to complex life. synthese 185: 21–52. ▶︎ http://cepa.info/2522

Schulkin J. (2011a) social allostasis: anticipatory regulation of the internal milieu. Frontiers in Evolutionary neuroscience 2: 3389.

Schulkin J. (2011b) adaptation and well-being: social allostasis. Cambridge university Press, Cambridge.

Seth A. K. (2013) interoceptive inference, emotion, and the embodied self. trends in Cognitive sciences 17(11): 565–573.

Seth A. K. (2014) The cybernetic Bayesian brain. in: Metzinger t. & Windt J. (eds.) open Mind. Mind Group, Frankfurt am Main: 35(t).

Steffensen S. V. (2011) Beyond mind. in: Cowley s. J. (ed.) distributed language. John Ben-jamins, Philadelphia Pa: 185–210.

Steffensen S. V. (2013) Human interactivity. in: Cowley s. J. & vallée-tourangeau F. (eds.) Cognition Beyond the Brain: Computation, interacivity, and Human artifice. springer, London: 195–223.

Steffensen S. V. (2015) distributed language and dialogism: notes on non-locality, sense-making and interactivity. Language sciences 50: 105–119.

Steffensen S. V. & Cowley S. (2010) signifying bodies and health. in: Cowley s. J. Major J. C., steffensen s. v. & dinis a. (eds.) signifying bodies: Biosemiosis, interaction and health. Portuguese Catholic university, Braga: 331–356.

Steffensen S. V. & Fill A. (2014) Ecolinguistics: The state of the art and future horizons. Language sciences 41: 6–25.

Steffensen S. V. & Pedersen S. B. (2014) temporal dynamics in human interaction. Cybernetics and Human Knowing 21(1–2): 80–97. ▶︎ http://cepa.info/2396

Steffensen S. V., Vallée-Tourangeau F. & Vallée-Tourangeau G. (2015) Cognitive events in a problem solving task: a qualitative method for investigating interactivity in the 17 animals problem. Journal of Cognitive Psychology 28(1): 79–105.

Steiner P. & Stewart J. (2009) From autonomy to heteronomy (and back): The enaction of so-cial life. Phenomenology and the Cognitive science 8: 527–550. ▶︎ http://cepa.info/476



http://www.ditext.com/mctaggart/time.html

http://www.ditext.com/mctaggart/time.html











Cogn

itiv

e SC

ienC

e Co

nCep

tS in

ena

Ctiv

iSm

264


Sterelny K. (2010) Minds: Extended or scaf-folded? Phenomenology and the Cognitive sciences 9: 465–481.

Sterelny K. (2012) The evolved apprentice. Mit Press, Cambridge Ma.

Sterling P. (2004) Principles of allostasis: optimal design, predictive regulation, pathophysiol-ogy, and rational therapeutics. in: schulkin J. (ed.) allostasis, homeostasis, and the costs of physiological adaptation. Cambridge univer-sity Press, Cambridge: 17–64.

Sterling P. (2012) allostasis: a model of predic-tive regulation. Physiology & Behavior 106(1): 5–15.

Stewart J., Gapenne O. & Di Paolo E. (eds.) (2010) Enaction. Mit Press, Cambridge Ma.

Thibault J. (2011) First-order languaging dynam-ics and second-order language: The distrib-uted language view. Ecological Psychology 23: 210–245.

Thompson E. (2004) Life and mind: From autopoiesis to neurophenomenology. a tribute to Francisco varela. Phenomenol-ogy and the Cognitive sciences 3: 381–398. ▶︎ http://cepa.info/1137

Thompson E. (2005) sensorimotor subjectivity and the enactive approach to experience. Phenomenology and the Cognitive sciences 4(4): 407–427. ▶︎ http://cepa.info/2358

Thompson E. (2007) Mind in life. Harvard uni-versity Press, Cambridge Ma.

Thompson E. (2011) reply to commentaries. Journal of Consciousness studies 18(5–6): 176–223. ▶︎ http://cepa.info/2347

Thompson E. & Stapleton M. (2009) Making sense of sense-making: reflections on enac-tive and extended mind theories. topoi 28: 23–30. ▶︎ http://cepa.info/2290

Torrance S. & Froese T. (2011) an inter-enactive approach to agency: Participatory sense-making, dynamics, and sociality. Humana Mente 15: 21–53. ▶︎ http://cepa.info/388

Turner J. S. (2000) The extended organism. Har-vard university Press, Cambridge Ma.

Turvey M. T. (2007) action and perception at the level of synergies. Human Movement science 26: 657–697.

Uryu M., Steffensen S. V. & Kramsch C. (2013) The ecology of intercultural interaction: timescales, temporal ranges and identity dynamics. Language sciences 41: 41–59.

van Orden G., Holden J. & Turvey M. T. (2003) self-organization of cognitive performance. Journal of Experimental Psychology: Gen-eral 132(3): 331–350.

Varela F. J. (1979) Principles of biological au-tonomy. Elsevier north Holland, new York.

Varela F. J. (1987) Laying down a path in walk-ing. in: Thompson W. i. (ed.) Gaia: a way of knowing. Political implications of the new biology. Lindisfarne Press, Hudson nY: 48–64. ▶︎ http://cepa.info/2069

Varela F. J. (1991) organism: a meshwork of selfless selves. in: tauber a. i. (ed.) organism and the origins of self. Kluwer, dordrecht: 79–107. ▶︎ http://cepa.info/1959

Varela F. J. (1992) autopoiesis and a biology of intentionality. in: McMullin B. (ed.) Proceedings of the workshop “autopoiesis an Perception.” dublin City university, dublin: 4–14. ▶︎ http://cepa.info/1274

Varela F. J. (1997) Patterns of life: intertwining identity and cognition. Brain and Cogni-tion 34: 72–87. ▶︎ http://cepa.info/2010

Varela F. J., Maturana H. R. & Uribe R. (1974) autopoiesis: The organization of living systems, its characterization and a model. Biosystems 5(4): 187–196. ▶︎ http://cepa.info/546

Varela F. J. & Shear J. (eds.) (1999) The view from within: First-person approaches to the study of consciousness. imprint academic, Thorverton uK.

Varela F. J., Thompson E. & Rosch E. (1991) The embodied mind. Mit Press, Cam-bridge Ma.

Vernon D., Lowe R., Thill S. & Zieme T. (2015) Embodied cognition and circular causality: on the role of constitutive autonomy in the reciprocal coupling of perception and action. Frontiers in Psychology 6: 1660. ▶︎ http://cepa.info/2523

Virgo N., Egbert M. D. & Froese T. (2011) The role of the spatial boundary in autopoiesis. in: Kampis G., Karsai i. & szathmáry E. (eds.) advances in artificial life: darwin meets von neumann. 10th European conference on artificial life (ECaL 2009) springer, Berlin: 234–241. ▶︎ http://cepa.info/2254

Weber A. & Varela F. J. (2002) Life after Kant: natural purposes and the autopoietic foundations of biological individuality. Phenomenology and the Cognitive sciences 1(2): 97–125. ▶︎ http://cepa.info/2087

Whitehead A. N. (1967) science and the mod-ern world. Free Press, new York.

Wilson A. D. & Golonka S. (2013) Embodied cognition is not what you think it is. Fron-tiers in Psychology 4: 58.

Wingfield J. C. (2004) allostatic load and life cycles: implications for neuroendocrine control mechanisms. in: schulkin J. (ed.) allostasis, homeostasis, and the costs of adaptation. Cambridge university Press, Cambridge: 302–342.















interactivity and enaction in human cognition ... · cognitive s c ien c e con c ept s in e na c...

Documents