the interplay of routine and decision making in · pdf filethe interplay of routine and...

25
1 The Interplay of Routine and Decision Making in Act Peter Kesting * First Draft, May 2007 Abstract: What precisely is the relation between routine and decision making? What opportunities do agents have to affect routines by decision making – and how does routine affect decision making in return? In this paper we will address these questions by understanding both, routines as well as decision making, as integrated elements of a higher entity: economic act. In the first section of the paper, we introduce our action framework. Following we outline, how routine can be acquired and applied in the course of repetition. As a result, we argue that – although of structurally different nature – both, routine and decision making, belong to the cognitive processes that link purpose and operations in act. Furthermore, we argue that with increasing repetitions, routine gains an increasing potential to substitute decision making. Finally we turn to the question of what opportunities actors have to influence routines by deliberate planning and decision making and how the acquired routine affects deliberate planning and decision making in return. We show that act indeed has to be understood as an interplay of routinezed repetition and deliberate planning and decision making. Keywords: Dynamic capabilities, routine, context changes, alertness, inertia. JEL-Classification: M13 * HHL-Leipzig Graduate School of Management, Department for Microeconomics and Information Systems, Jahnallee 59, 04109 Leipzig, Germany, email: [email protected] ..

Upload: duongnguyet

Post on 05-Mar-2018

219 views

Category:

Documents


2 download

TRANSCRIPT

1

The Interplay of Routine and Decision Making in Act

Peter Kesting*

First Draft, May 2007

Abstract:

What precisely is the relation between routine and decision making? What opportunities do

agents have to affect routines by decision making – and how does routine affect decision

making in return? In this paper we will address these questions by understanding both,

routines as well as decision making, as integrated elements of a higher entity: economic act.

In the first section of the paper, we introduce our action framework. Following we outline,

how routine can be acquired and applied in the course of repetition. As a result, we argue that

– although of structurally different nature – both, routine and decision making, belong to the

cognitive processes that link purpose and operations in act. Furthermore, we argue that with

increasing repetitions, routine gains an increasing potential to substitute decision making.

Finally we turn to the question of what opportunities actors have to influence routines by

deliberate planning and decision making and how the acquired routine affects deliberate

planning and decision making in return. We show that act indeed has to be understood as an

interplay of routinezed repetition and deliberate planning and decision making.

Keywords: Dynamic capabilities, routine, context changes, alertness, inertia.

JEL-Classification: M13

* HHL-Leipzig Graduate School of Management, Department for Microeconomics and Information Systems,

Jahnallee 59, 04109 Leipzig, Germany, email: [email protected]..

2

The Interplay of Routine and Decision Making in Act

INTRODUCTION

Currently, there are two different approaches to economic act: Routine and (rational) decision

making. These two approaches fundamentally differ regarding their basic assumptions as well

as their results. The rational decision making approach understands economic act as an

outcome of deliberate planning, a procedure in that economic agents deliberately identify and

compare their alternatives and chose the best one according to their preferences. Abstracting

from any bounds of human rationality, this approach comes to the result that economic act

always incorporates optimal solutions for given problem settings (Samuelson 1983, Jeffrey

1983). In contrast, the routine approach stresses the bounds of human rationality and assumes

that economic act is not a result of a planning procedure – at least not in the traditional sense

–, but consists of ‘behaviour that is learned, highly patterned, repetitious, or quasi-repetitious,

[and] founded in part in tacit knowledge’ (Winter 2003: 991). Consequently, economic act is

not always fullys adapted to the context and hence not optimal in a traditional, substantial

sense. As a result, we have two competing approaches that explain one and the same thing in

a different way and come to different results.

How to deal with that situation? Do the different approaches fundamentally exclude each

other? Or do they complement each other? And if they complement each other, in what way?

What precisely is the relation between routine and decision?

It appears not exaggerate to us to state that a majority of exponents of both approaches

have chosen to deal with the situation by simply ignoring each other. For the decision making

approach, ignorance of routine is somewhat structurally given. As we will argue below,

routine, first of all, is a strategy to deal with the human bounds of rationality. By assuming

humans as being substantially rational, the rational decision making approach leaves no room

to fill the concept of routine with a meaningful content.

For the routine approach, the situation is a little more subtle. Stemming from the bounded

rational approach, the basic concept of routine does not principally exclude deliberate

planning and decision making from its explanation – at least not from the outset. However,

the routine approach has been developed in the consequence of a critique of strong rationality

3

assumptions of the decision making approach and explicitly aims to challenge this approach.

From the beginning, it was presented as an alternative. Nelson/Winter’s statement ‘the

behavior of firms can be explained by the routines that they employ. Knowledge of the

routines is the heart of understanding behaviour. Modelling firms means modelling the

routines and how they change over time” (1982: 128), is undeniably a programmatic and

offensive one. After all it says nothing more than that the traditional, decision making

approach has almost no explanation power to understand firm behavior. And indeed, in the

few conceptual works on routine, deliberate planning and decision making are only of

secondary importance – if at all (in particular Simon/March 1958, March 1995, Cohen et al.

1996, Feldman 2000 and 2003, Winter 2003, Pentland/Feldman 2005, Becker 2005). Winter

(2003) does not even employ decision making to explain the creation of routines at all, but

super-ordinate, so-called second-order routines.

After all, this is not very surprising. We already know from Thomas S. Kuhn (1970) that

when challenging an established paradigm, ‘revolutionaries’ have a tendency to create and

overemphasize a distance. For the routine approach, this naturally means to play down the

meaning of rationality for the explanation of act. However, the separation between routine

and decision making became ultimately manifested by defining routine as ‘pattern of activity’

(Dosi, Teece und Winter 1992, Dosi, Nelson und Winter 2000, Feldman 2000, Feldman und

Rafaeli 2002), i.e. as an independent class of act and almost insurmountably by defining

routine as an organizational phenomenon exclusively (Hodgeson 2003, Becker 2005).

As a result, there are almost no systematic investigations into the relation between routine

and decision from either side with the consequence that we still know very little about this

issue. However, we are strongly convinced that this is wrong and that our ignorance results in

a serious deficiency. We are convinced that both, routine as well as decisions are important

elements of human act and that we have to find a synthesis that explains human act as an

interplay of both.

What is the relation between routine and decision? What opportunities do individuals or

organizations have to control their routines by deliberate planning and decision making – and

how does the acquisition and application of routine affects deliberate planning and decision

making in return? In this paper we will address these questions by understanding both,

routines as well as deliberate planning and decision making, as integrated elements of a

higher entity: economic act. Hence, to pave the ground for the investigation of our main

4

problem we first introduce our action framework in the first section of this paper. We require

this action framework first of all to define the entity of one act and – based on that – a concept

of repetition that is independent of routine. This independent concept of repetition is a

necessary precondition for our investigation of the gradual acquisition and application of

routine in the course of repetition in the following section. As a result, we argue that –

although of fundamentally different nature – both, routine and deliberate planning and

decision making serve the same purpose: to determine what operations to conduct to bring

about a certain intended state. For this reason, routine has the potential to substitute decision

making. Against this background we intensively discuss the question of how to define the

notion of routine most appropriately. Finally we investigate how the acquisition of routine

changes decision structures in the course of repetition and how to affect the ‘stock’ of routines

by decision making in return. We argue that act indeed has to be understood as an interplay of

routine and deliberate planning and decision making.

THE BASIC CONCEPT OF ECONOMIC ACT

Both, routines as well as decisions can be understood (and have been understood many times)

as entities on their own. However, the basic approach of this paper is to understand both –

routine as well as decision making – as elements of a higher entity: human act. This means in

particular that human act is more comprehensive and contains more elements than routine and

decision making alone. (It also means that there is no routine and decision making outside

human act, but this is not critical for this paper.) The strategy of this paper is to establish a

link between routine and decision making by identifying the function that they serve for

action taking. For convenience, we will label this procedure as ‘action-based approach.’

Unfortunately, we do not have an ‘independent’1 concept of act in economics. For this

reason, we have to fall back on philosophical action theory. What is an act? Philosophical

action theory defines act as ‘purposeful behavior’ (Anscombe 1954, Davidson 1963, v.

Wright 1971). Against this background, the entity of one act can be defined as the triple

[intention, operations, deliberation]: An act consists of an intention that defines a specific

state that an agent desires to bring about. It additionally includes all deliberation and all

(physical) operation that an agent undertakes (and only this) to bring about the intended

state. This concept requires some further elaboration. 1 In this connection, independent means that act is not immediately explained as an outcome of a decision or

identified with ‘a routine,’ but defined in terms genuinely independent of routine and decision making.

5

(1) The first element of an act is the intention. For economists, this element is perhaps most

unfamiliar since traditional microeconomics as well bounded rational approaches both

represent human pursuit by preferences alone. There has been an intensive and rich discussion

in philosophical action theory about purpose and intentionality in act, but in economics and

decision theory similar concepts has been discussed only recently and only by very few

researchers (van Hees 2003, Roy 2006, Kesting 2006). We define an intention as a desire to

bring about a particular state within a certain time frame. State defines one and only one

characteristic of a future world and does not define the future world as a whole (i.e. all

characteristics of it in detail). In this sense, an intention does nothing more than to define a

specific target an agent wants to attain in the future. One example for such an intention would

be to give the Introduction to Entrepreneurship at the University X in the summer term Y.

Intentions – like other goals – can form a hierarchical structure. There can be sub- or super

ordinate intentions. For instance: To give a class on entrepreneurial spirit can be quite an

intention on its own, but at the same time, it can be part of the super-ordinate intention to give

the Introduction to Entrepreneurship. However, there is always a ‘lowest’ hierarchical level

of intentions with no sub-ordinate intentions.2

Precondition of the formulation of an intention is that agents have to be convinced that the

intended state does not establish itself ‘by alone,’ but that they have to do something to bring

it about actively (within a desired time frame). As long as this is not the case, agents don’t

have to act, but just to wait. Consequently, the main function of the intention is to trigger and

define the agents’ activity. As long as this is not the case, the intention remains at the level of

unspecific desire.

Do the agents actually have to be convinced to be capable of bringing about the intended

state? Not necessarily. One can also imagine that an intention triggers further activity even

when agents have no idea of how to bring about the intended state and even when they doubt

that they can ever do it. In this case, the agents begin to strive for reaching an intended state.

However, striving is quite different from ‘ordinary’ decision making and we don’t want to go

into it in detail in this paper.

Against this background, it appears necessary to us to emphasize that intentions denote a

fundamentally different thing than preferences and should not be mixed up with these. 2 For simplification, we will assume that every subordinated intention serves one and only one superordinated

intention.

6

Intentions are targets that trigger further activity. In contrast, preferences denote the ability to

value or rank different alternatives. They are a necessary precondition for rational decision

making. Preferences do not trigger any activity and can also not be satisfied. Of course, there

is some kind of relation between these different expressions of human pursuit. But this not

relevant for the analysis of this paper (for an closer investigation of the relation between

preferences and intentions see Kesting 2006).

(2) The second element of every act are the agents’ operations. Basically, operations denote

all (physical) activities – and only these – that are carried out in order to bring about the

respective intended state.3 Against this background, intentions and operations form a goal –

mean relation.

It are these physical operations that distinguish act from desire; act always includes

activity, i.e. a systematic and purposeful interference with the ‘way of the world.’ Physical

operations include verbal communication since there are many cases for that verbal

communication is necessary and even sufficient to bring about an intended state. One example

for this is our Introduction to Entrepreneurship which almost completely consists of verbal

communication.

(3) If the intention does not remain at the unspecific level of desire but indeed triggers the

agents’ activity, it necessarily raises the question of what operations to conduct to bring about

the intended state. The intention alone does not determine operation. It only determines one

specific state to be brought about – not less, but also not more. It does not determine how to

bring about the intended state, i.e. how to go from here to there. And it does not determine

how the future world looks like apart from the realization of the intended state. As a

consequence, agents usually4 have different feasible options or ‘alternatives’ to realize the

intended state. In our example, the intention determines the target to give a lecture and

possibly makes some additional specifications like number of credit points or the time frame

of the lecture. But it does not determine the content, structure, teaching methods etc. of the

lecture.

3 As Donald Davidson (1971, p. 23) has put it so fabulously: ‘We only can use our bodies. The rest is up to

nature.’ 4 It is also possible that an agent determines only one alternative to bring about a certain intended state. In this

case, decision making is – of course – not necessary.

7

This is the point where the third element of action comes into the game: deliberation. We

define deliberation as all activities that an agent pursues to answer the question of what

operations to conduct in order (and only in order) to bring about the intended state. This

means that deliberation includes decision making, but in contrast to the traditional view, it

consists of much more than that. It also includes all pre-decision making activities, in

particular all information acquisition and processing in order to identify and characterize

feasible alternatives to bring about the intended state (which ultimately constitute the decision

problems which are usually assumed by traditional investigations as given). Moreover,

deliberation also includes all formal documentation, adjustment and application of routine.5

Before carrying on we should devote some additional some words to the basic function of our

action framework and its contribution to an analysis of the interplay of routine and decision in

act. For economic analysis, not only the intention, but also the conceptualization of

deliberation as connecting element between intention and operation is uncommon. As we will

argue below, this conceptualization of deliberation is much broader than the traditional

decision making approach.

The main function of intentions is to structure act. Basically, human activity forms a

continuous flow of operations. Intentions suit the purpose to define an entity of one act within

this continuum: One act consists of an intention and of all deliberation and all operations an

agent undertakes to bring about this specific intended state. Consequently, also the complexity

of an act is completely determined by its intention. It can be as complex as the act of

‘bringing a man to the moon’ or as simple as the act of ‘opening a window.’

The justification of the employment of intentions for economic analysis is a complex

matter that goes far beyond this paper.6 Certainly, to structure economic activity into a

number of different somewhat independent acts is artificial, although this is already common

practice in traditional decision theory. There is also a number of cases for which the

explanation power of intentions (and with it the scope of the action-based approach) might be 5 Although quite straight forward in principle, the distinction between operation and deliberation is far from

being trivial. Basically, both operations as well as deliberation denote classes of human activity (cognitive and

physical). The principle difference is that operations include all activities that in fact change the world in the

desired direction and deliberations all activities that are carried out in order to determine the operations.

However, we are aware that it is very difficult to specify this distinction for particular cases. But this problem is

not crucial for the investigation of this paper. 6 For detailed discussion see Kesting 2006.

8

challenged.7 However, the analysis of this paper builds on this concept in particular for two

reasons:

First, intentions constitute a very effective strategy to reduce complexity and focus de-

liberation. Once the intention is given, planning can be reduced to one and only one problem:

What operations to conduct in order to bring about a certain intended state? Everything that is

relevant to solve this question belongs to the decision problem, everything that is not relevant,

can be excluded from planning. Technically spoken, intentions are suitable to define small

world decision problems with a well defined set of feasible alternatives that covers only a

(very small) subset of the totality of the possibilities to act an agent has at a certain point in

time (Savage 1954, Kesting 2006). There are many indications that the complexity of reality

completely overwhelms human deliberation. Against this background, the formulation of

small world decision problems is a vital a precondition for rational decision making at all. Via

the intention, agents can immediately determine the complexity of a decision problem.

Additionally, intentions serve the purpose of guiding (and focusing) the acquisition of

information and the search for alternatives within in the pre-decision stage. After all,

intentions – like routine – are hence nothing more than a strategy to deal with the human

bounds of rationality. Small world decision problems are never optimal in a substantial sense

since they systematically neglect their connection to the grand world.

Second, and much more important for this paper, intentions allow to qualify what it means

to repeat an act: To repeat an act means for the same agent to bring about the same intended

state a second or repeated time in a somewhat similar context (within a certain time frame,

given somewhat stable preferences). This conceptualization of repetition underlies the entire

analysis of this paper. It will turn out that repetition is a precondition for the acquisition as

well as the application of routine. Against this background, the meaning of the entire

conceptualization of act becomes obvious: to describe the setting in that routine can be

acquired and applied. Current routine research does not offer any closer specification of 7 This is in particular the case for reflexes and for activities that are driven by intuition and do not follow an

explicit intention. But do we really develop routine for these kinds of activity? More serious is the case of

imitation. Even complex activities can be determined without any intention, simply by imitating the behaviour of

the parents, colleagues, competitors or others. And it is certainly possible to develop routine for such activities.

One can argue that it is not very rational to do so and that it is imperative for our act to follow a certain purpose

even when it is a result of imitation. However, in cases where this is not so, the explanation power of the action-

based approach is rather restricted.

9

repetition independent of routines. We are convinced that this deficiency causes serious

problems for the foundation of the concept, in particular regarding the question of how

routine is acquired and applied, how changeable routine is and inasmuch it resists towards

changes in the course of repetition and – ultimately – how to define routine.

THE GRADUAL SUBSTITUTION OF DECISION MAKING BY REPLICATION IN

THE COURSE OF REPETITION

In this section, we focus on the question of what opportunities agents have to determine what

operations to conduct to bring about a certain intended state and how these opportunities

change in the course of repetition. We will start with ‘the first act,’ proceed to the first

repetitions and end up at the highly repetitive ‘best practice.’ We will argue that repetition

leads to a gradual substitution of deliberate planning and decision making by replication.

However, we will refrain from introducing the notion of ‘routine’ already in this section.

We will focus on repetition and its associated processes. We will discuss why and how action

taking changes in the course of repetition. Based on this investigation, we will turn to the

question of how to define routine in the subsequent section.

The first act

At some point in time, every act is taken the first time – it necessarily has to. Now, there is

general agreement among routine researchers that routine is acquired via repetition (this will

be also one of the results of this paper). Yet, this implies that when taking ‘the first act,’

agents have not the opportunity to fall back to their routine but have to solve the problem of

what operations to conduct to bring about a certain intended state somehow differently. So

how to determine ‘the first act’? It appears a little as if routine research somewhat would duck

out this question. There are no clear statements and no substantial discussions regarding this

point. We are left only with a few remarks, in particular by Nelson/Winter (1982) that argue

that the first acts are determined by ‘trial and error.’

The classical answer to this question – of course – is that agents solve this problem by

rational decision making. Ultimately, this implies that agents are always capable to find and

realize the best alternative they have (in terms of their preferences) to determine their

operations. The only key to efficiency in act is hence proper decision making.

There have been many discussions about the explanation power of this extreme version of

economic rationality. In this paper, we take up a somewhat ambiguous position to it.

10

On one hand, it can hardly be denied that humans have more opportunities to determine

what operations to conduct to bring about a certain intended state than decision making alone

– even for the ‘first act.’ Alternatives include intuition, imitation, and instincts.8 However, the

importance of decision making for human act and with it the explanation power of the

traditional decision-approach can also hardly be denied. At least two reasons speak for a

traditional, decision-based explanation of the first act: First, there is strong evidence for that

humans indeed actively search for the best alternative to carry out an act, although the effort

that is devoted to this search and the success of this search certainly considerably varies.

Additionally, it is completely implausible to assume that humans act irrationally in the way

that they deliberately opt for a sub-optimal alternative. Second, deliberate decision making is

the strongest mean humans have to make act purposeful. Decision making is nothing more

than an active and deliberate search for the best way to act. If humans do not make use of this

opportunity and fall back intuition, imitation, or instincts, they miss an opportunity to make

act purposeful. For these reasons we will focus our explanation of the first act on rational

decision making (keeping in mind that alternative options are existent). At least it can hardly

be denied that humans indeed have an opportunity to determine the first act by deliberate

planning and decision making.

On the other hand, however, we fundamentally disagree with the classical assumptions

regarding the acquisition and processing of information. These strong assumptions block any

investigation of the imperfections which are crucial to understand human act – in particular

when analyzing the first, non-routinized act in that agents face a new situation that they can

address only on base of their unspecific background knowledge (Runde 2002) or by acquiring

new information which is costly and time consuming. The most relevant imperfections are

that when facing a new and somewhat complex decision problem, agents are usually not able

(i) to identify all alternatives they have, (ii) to describe all identified alternatives accurately in

every detail, and (iii) to conclude to the outcome and hence to value the alternatives correctly.

As a consequence, when taking an act for the first time, agents are usually not able to identify

the best alternative they have. We assume that agents indeed decide rational, but only in the

sense that they ‘strive’ to identify and chose the best alternative they have. The result of this

activity is necessarily constrained by the human limitations to acquire and process

8 We do not include reflexes into the list since we doubt that humans can develop routine for these. That humans

can not acquire routine for reflexes is an interesting fact on its own that points out the conscious dimension of

routine.

11

information. As a result, agents are systematically not (or only coincidentally) able to

establish a substantial optimum (in the classical sense) when taking a new act the first time.

With this assumption, this paper is perhaps goes along with a majority of the so called

bounded rationality literature (Conlisk 1996).

However, the decision of what operations to conduct to bring about the intended state does

not terminate the act. Indeed, the determination is followed the operations, i.e. by the physical

realization of the intended state. The act is only terminated when the intended state is brought

about.

This ‘post-decision phase’ is particularly important because it provides an opportunity for

agents to observe the process and acquire specific knowledge about the act. In particular, they

can learn all the unplanned details of an act. Almost always, planning is incomplete in the

sense that it does not cover every detail of an act. When planning to travel from A to B the

first time, we plan what mean of transportation to chose and on what route to travel. But we

do not plan every step we take. We rely on that we are capable to manage the situation once

we are in it. However, this includes that we usually encounter unexpected difficulties that

cause additional planning effort during operations and often even force us to revise our

original plan. But most of all, in the post-decision phase we have the opportunity to observe

the outcome of our operations. We can observe, if our expectations go along with the reality

or not. This gives us a completely new base to assess the alternative we have opted for.

As a result, by taking an act the first time, agents have the opportunity to improve their

specific knowledge about the question of what operations to conduct to bring about a specific

state. Agents increase their knowledge in the pre-decision phase when they identify and

characterize alternatives, in the decision phase when they compare different alternatives and

make a decision and in the post-decision phase when they can observe the course and the

outcome of their operations.

Finally, we can sum up the two key results of this section: (1) Decision making is not the only

option that humans have to determine the ‘first act,’ but it certainly is an option – and a very

powerful one. (2) Due to the existing bounds of rationality, planning is necessarily incomplete

and does not (or only accidentally) lead to an optimal result (in the traditional, substantial

sense). The same is certainly true for intuition, instincts, or imitation.

12

Repetition

It is generally acknowledged among routine researchers the repetition plays a central role for

both, the acquisition and the application of routine (e.g. Winter 1986, Eisenhardt und Martin

2000, Pentland 2003a, 2003b, Becker 2005). But how do the underlying processes precisely

look like? What does acquisition and application of routine precisely mean? Routine research

does not present any clear answer to this question, mainly because it does not offer an

independent concept of repetition so far. Instead, it often employs repetitiveness an a

definition criteria, which makes the entire concept somewhat circular. (If routine is defined by

repetitiveness, how can repetition – which is certainly repetitive – be the basis to acquire and

apply routine.)

In this section, we investigate what happens when agents repeat an act. In particular, we

focus the analysis on deliberation and the gradual substitution of planning and decision

making by replication in the course of repetition. At this stage, we do not introduce any

definition of routine. Instead, we will built on the results of this section to discuss how to

define the notion of routine most appropriately in the subsequent section.

We remember: To repeat an act means for the same person to bring about the same intended

state again a second or repeated time in a somewhat similar context. Per definition the

‘objective’ problem setting remains the same as before. However, it is the agents that change

since they now already have solved the attached decision problem and acquired specific

knowledge about this very act (which they did not have before).

Every execution of an act incorporates a comprehensive answer to the problem of what

operations to conduct in order to bring about the intended state (including even those aspects

that were not included into previous planning and decision making and only have been

determined ‘spontaneously’ during operation) (see also Betsch et al. 1999, Betsch 2005). It

necessarily has to, since agents actually take the act. We will call this comprehensive answer

a solution.

However, at complex acts, these solutions are complex as well. They do not only consist

of an overall concept, but have many different characteristics regarding a multitude of

different ‘aspects’ and ‘sub-aspects.’ In our lecture example, the solution includes an overall

agenda and an overall didactic approach. But this overall concept has to be filled with a

multitude of details. It has to be determined what entrepreneurship frameworks to discuss and

how, what readings to request and what readings to recommend, what case studies to use for

13

the different topics what transparencies to draw and how and many more things.

Keystone of all changes in the course of repetition is specific learning. When taking the act

the first time, the agents already have identified, specified and valued alternatives. Moreover,

they experience the realization of the act. They can observe formerly unplanned details and

difficulties they encounter during execution. And they can observe the result of the act. This

learning is specific since it regards the question of what operations to conduct to bring about a

certain intended state. The additional knowledge is exclusively knowledge of one particular

solution and its various sub-solutions.

This kind of learning has been very well acknowledged by traditional decision theory.

However, what has not been acknowledged so far is the particular meaning of this additional

knowledge for agents when their rationality is bounded and planning causes (opportunity)

costs. If this is the case, it is only rational for agents not to go through the entire decision

making procedure again with every repetition, but to apply previous solutions or sub-solutions

in case they are content with these. Agents are content with a previous solution when they do

not expect that additional planning (including the acquisition of additional information) pays

off, i.e. when they do not expect that the improvements outweigh the ‘costs’ of additional

planning. In this case, they still make a decision. But the decision structure is ‘degenerated’

since agents shortcut the entire decision making process and reduce the entire planning to the

pure decision to ‘do it like last time.’ We will phrase this application of a solution as

replication. In case they are not content with the result of a previous solution, it is only

rational for them to carry on planning. In this case, they carry on identifying, specifying and

valuing alternatives and make a ‘real’ decision. The main difference to the traditional view is

that planning is now incomplete since solutions or sub-solutions are taken over unquestioned

‘as a block.’

However, it is important that this re-assessment of previous solutions, again, does not only

concern the overall conception of the act, but also all its aspects and sub-aspects. To return to

our example: After giving the Introduction to Entrepreneurship the first time, we might

question the overall didactic approach and several details of the lecture, but we might be

content with some of the case studies and many of the transparencies that we have used. As a

result, we might revise our solution fundamentally, but not completely. Already at the first

repetition we replicate some aspects of the first solution. For instance, we employ the case

study on entrepreneurial finance without further deciding about it, i.e. without searching for

additional alternatives anymore.

14

With increasing repetitions, agents tend to find an increasing number of solutions to

aspects that they are content with so that the replicated part of the act tends to increase in the

course of repetition. It follows: The substitution of decision making by replication is a

gradual process and not an all or nothing. In our example, at some point in time, we might be

content with the overall concept and the textbooks we employed. In the course of the next

repetitions, we confine ourselves to revise only details and that even to a decreasing extent.

Ultimately, we tend to find a satisfying solution which cover more or less every aspect and

sub-aspect. In the following, we will phrase this comprehensive program as ‘best practice.’9

There is some evidence that with increasing repetitions, acts increasingly become sub-

conscious and automated (Winter 1986, Nightingale 2003). At complex acts, this particularly

concerns the ‘course of the act.’ Automation makes agents particularly capable of ‘uncoiling’

an act automatically without further deliberation. One example for that is that when driving a

new route, we constantly have to take a look at the road map. After driving the same route a

several times we do not have to consult the map anymore, but find the way ‘automatically.’

This in particular saves deliberation effort in the post-decision phase. This and the

simplification of planning and decision making procedures brings about that by replication

agents can save a considerable amount of deliberation effort. This makes replication a

powerful strategy to deal with the human bounds of rationality.

It has to be emphasized that we can only acquire and apply solutions as long as intention and

context remain stable. If only one of these change considerably, the ‘objective’ problem

setting changes and there is no systematic reason why a previous acquired solution should

produce satisfying results anymore. Against this background, the entire conceptualization of

repetition becomes clear: it describes the necessary and sufficient conditions to acquire and

apply solutions.

But if intention and context remain stable, and as long as agents remain rational (and once

acquired knowledge does not fall into oblivion) there is only one overall direction for change:

improvement. The reason for that is very simple: Agents would never deliberately fall back

below a previous solution, but increase their knowledge with every repetition and actively

search for improvements. This makes agents to become ‘familiar’ with an act not only in the

9 It is possible that repetitions run out of purpose in the sense that they do not follow any purpose anymore, but

are carried out in certain situations habitually. In this case they become irrational when agents do not want to

bring about the intended state anymore.

15

sense that they know one or more solutions for different aspects, but that they learn about all

the details of the act. The only reasons to fall back behind previous results are mistakes, bad

luck or the failure of a trial. However, must be expected that this improvement does not lead

to an substantial optimum (in the traditional sense) when rationality is bounded and when

agents are already content with a sub-optimal result.10

Seen in this light, repetition is not a classical learning circle for at least two reasons: First,

to reply a solution per definition means to apply it without further planning. As a result, the

respective aspect of the act is repeated unchanged. Any changes result from additional

planning (or improvisation). As a consequence, the feedback is partly cut and any additional

information regarding this aspect is ignored. Second, and in the same direction, also

automation leads to an identical repetition and to a systematic ignorance of information. Both

reasons cause that repetitions become patterned and much more inflexible as suggested by

Feldman (2000). They causes that repetitions are not always adapted to context changes. For

this reason, repetitions always run in danger to become ineffective.

However – already Nelson/Winter (1982) have pointed out that agents always have the

opportunity to question their best practice and re-plan an act. And here, decision making

comes into the game again. Basically, agents have two different opportunities:

• Either they can take a completely new, non-repeated act, i.e. an act that follows a new

intention which is not sub-ordinate to that of an established act. In this case, the agents

would act outside the logic of the antecedent acts. They would do some-thing

substantially new. 10 There is obviously some relationship between this comprehensive best practice and the neoclassical optimum.

But this relationship is not a simple one. Very similar to the neoclassical optimum, also the best practice de-

scribes a situation that is very well mastered by the agent and that tends to be solved efficiently. In this sense,

one could argue that agents act as if they have solved the respective decision problem. And in this sense one can

argue (and indeed, we share this position) that traditional optimisation provides a somewhat accurate description

of the continuation of an act in a somewhat stable context, for instance of the circular flow of that kind as de-

scribed in the opening section of Schumpeter’s Theorie der Wirtschaftlichen Entwicklung (1935, Ch. 2). In such

a situation, even the action-based approach predicts a tendency towards equilibrium.

However, it has to be emphasized that according to the action-based approach, the driving force for effi-

ciency is not substantial rationality (like assumed by the traditional approach), but familiarity with the problem

setting, resulting from learning, trial and error, and gradual improvements in the course of repetition. This differ-

ence becomes particularly efficient when agents are forced to act outside their best practice in a changing con-

text. Here, results are far from being efficient and the traditional approach only provides a poor approximation to

reality. We will address this problem in the next section.

16

• Or the agents can re-plan an old act and choose a new way to repeat it. In this case,

they question and abandon an established best practice. But they do not leave the logic

of the established act, meaning they do not base their deliberation and operation on a

new intention. They just search for a new way to do an old thing.

Basically, breaking with the established routine somewhat brings actors back to the ‘first act.’

When taking a new act, agents can not built on best practice since best practice is always

acquired in a specific context to bring about one particular intended state and only fits to this.

There is no systematic reason that solutions, which lead to a satisfactory outcome to bring

about one particular intended state also lead to a satisfactory outcome to bring about a

different intended state. When re-planning an act, agents do not want to built on old best

practice but look for a new, better way to take an act. In both cases, agents necessarily lose

the advantages of replication and best practice, in particular:

• Planning effort significantly increases since every new act has to be planned from the

scratch (Nelson/Winter 1982).

• Efficiency decreases (at the beginning) since at every a new act, agents per definition

start at the low end of the learning curve (Arrow 1962).

• Uncertainty substantially increases since when taking an act a first time, agents can

not built on previous observations and can only conclude to its outcome

(Becker/Knudsen 2004).

As a result, the decision between breaking with a best practice or continuing it has an

asymmetric structure: It compares a familiar, somewhat certain and efficient solution with an

‘adventure’ which is risky, inefficient, and causes a lot of planning effort (in a world in that

time is precious). As a result, agents have a tendency to lock-in to existing routines.

We do not want to go too deep into the analysis of this phenomenon. It might just be

mentioned that this lock-in is partly rational since it follows a simple discounted cash-flow

logic that determines the net-effect of change. Lock-in means in this connection not to replace

solutions even when there are alternatives that have a supposedly higher efficiency outcome.

However, it is not only difficult – if not impossible – for bounded rational actors to determine

the net-effect of replacement. Partly, however, this ‘lock-in’ is also irrational and results from

cognitive biases, from incentive structures and from resistance towards change (see Kesting

2006). In the following, we will phrase this ‘lock-in’ by characterizing a solutions and best

17

practice as ‘established.’11

Let us summarize the results of this section: In the course of repetition, deliberation and

operation of an act change. Engine of this change is replication, i.e. the application of former

solutions to current problems. Such an application is only possible if context and intention do

not change significantly. Following, intention has to be an essential element of every

investigation of the acquisition and application of solutions.

Humans have the opportunity to break from their established solutions and best practice:

They can do acts new or they can do new acts. But the underlying decision to do so is

asymmetric since it compares the familiar with the unknown. As a result, agents have a

rational and irrational disposition to stick witch established solutions.

HOW TO DEFINE ROUTINE?

By now there is no final consensus on how to define routine. However, in this section we

argue that the situation is even worse. We argue that all existing definitions are imprecise and

leave crucial aspects undetermined. Moreover, they do not progress to the underlying driving

forces that make agents to acquire and apply routine, but remain at the (empirical) surface of

their mere appearance. For this reasons, these definitions do not suit as a solid foundation to

approach the underlying phenomena. (Perhaps this is the main reason that no common sense

had been reached so far.) And in particular they are not sufficient to challenge traditional

decision theory. In this section we discuss the three most common approaches to define

routine: pattern of activity, rule, and mental representation.

(1) Currently, an increasing number of researchers appears to understand and conceptualize

‘routines’ in accordance with Cohen et al. (1996) as patterns of activity (Winter 1986,

Teece/Pisano/Shuen 1997, Feldman 2000). However, it is not simple to ‘translate’ this

conceptualization into the terms of our action framework, in particular because basic elements

of this definition have never been closer specified by their proponents.

11 It has to be noted that it is also possible that repeated acts run out of purpose (see in particular Nelson/Winter

1982). In this case, act is not triggered by the formulation of an intention anymore, but ‘behaviour’ by a stimulus

(we use the word behaviour at this since we have defined act as ‘purposeful behaviour’ that follows an

intention). The response to such a stimulus takes place automated and the purpose of act is not consciously

reflected anymore. I would propose to call this kind of behaviour a ‘habit.’

18

Let us start with the second element of the definition, the ‘activity.’ There is no closer

specification of what activity precisely means. However, we can see from our action

framework the term activity is not unambiguous and can mean three qualitatively different

things: intention setting, deliberation and operation. Hence, we are up to an interpretation at

this point. We guess that it is most plausible to understand activity comprehensively and not

to limit it on one specific subset (e.g. operations). In this translation, routines are identical

with acts (in our understanding).

‘Pattern’ is obviously used as distinctive criterion: Routines do not denote all acts, but a

subset, namely all those acts that are patterned. But what does patterned precisely mean?

Certainly, it means similar, but not identical. But how to specify similarity? This is far from

being trivial and becomes crucial when similarity is employed as distinctive criterion of

routine. First, any definition has to specify the objects of similarity. What is similar to what?

Here the most plausible answer is that the objects are different acts and the similarity is the

similarity of repetition. For this reason, the first act can never be a routine. But at what time

do repetitions cross the threshold of similarity? As argued above, acts can even fundamentally

change at the first repetitions. Nevertheless, they still follow the same purpose in the same

context. This appears to be the most fundamental similarity – the most significant repetitive

jump is that from the first act to repetition. For this reason, it appears most plausible and

straight forward to specify patterned by repetition. In this case, ‘routines’ are nothing more

than what we have denoted as repeated acts.

However, such a definition does not really suit as starting point for an investigation of the

phenomenon for at least three reasons:

First, according to this definition, ‘routines’ would include decision making procedures.

Decision making would become a part of routines. As argued above, decision making

continues at repetitions of an act. Decisions might be less comprehensive, but still they are

there and play a role to determine the ‘routine.’ This does not only break with our customs to

use the word. Most of all, it prevents to understand routine and decision as antipodes. The

antipode to routines are now – the first acts.

Second, this definition conceptualizes routines as a once-for-all phenomenon: an act is a

routine or it is not, there is nothing in between. This prevents ‘routine’ from the outset to

serve as a foundation for the investigation of all gradual processes that take place in the

course of repetition (like learning or the gradual substitution of decision making by

replication) – it makes routine blind for these processes. It conceptualizes a gradual process

19

by a qualitative jump.

Third, and perhaps most important, the pattern definition categorizes routines at the

surface of their mere appearance as a specific class of acts and does not progress to the

cognitive processes below this surface. This reminds a little to times when diseases were

conceptualized by their symptoms and the understanding of diseases did not progress to the

driving forces behind these symptoms: bacteria and viruses. We are convinced that every

theory of routine has explain the formation of patterns, but not to start with it. For this reason,

it has to address the cognitive processes that determine act – consequently, it has to

conceptualize routine as an ‘cognitive alternative’ to decision making.

To conclude, we might express that we do not want to question the fact itself: We share

the conviction that ‘routines’ indeed form patterns in the course of repetition and that the

formation of patterns is an important characteristic of routines. What we want to question,

however, is the suitability of pattern as definiens of routine and as starting point for the

investigation of the underlying phenomenon.

(2) As an alternative, routines has also sometimes been defined as rules. Since this approach

was not very influential for routine research we want to keep the discussion short and only

address the most critical point. Cambridge Advanced Learners Dictionary (2003) defines a

‘rule’ as ‘an accepted principle or instruction that states the way things are or should be done,

and tells you what you are allowed or are not allowed to do.’ In our action framework such

rules can only be identified at highly codified best practice. Such a definition would

fundamentally limit the meaning and explanation power of routine for action taking. As a

consequence, routine would become an aspect of act, but not a fundamental cognitive

principle. Moreover, routine would not serve as conceptual foundation for the investigation of

the effects of repetition.

(3) A third approach is to define routine on the cognitive level, usually as a ‘mental

representation’ of standard answers in the sense of a ‘program.’ This definition goes into the

right direction since it addresses the mechanisms that drive the formation of patterns.

However, it requires a very careful specification since the underlying cognitive processes of

repetition are many and diverse. They include:

• The determination of an intention.

• The entire planning procedure prior to operation, namely the identification,

20

specification and valuation of alternatives and the decision which results in a program

or plan that determines how to take the act.

• Cognitive processes that underlie operations, in particular to recall and adapt the

decision, meaning to improvise and solve additional problems where this is necessary.

• Finally the entire acquisition of specific knowledge in the course of repetition as a

result of observations and experience. This knowledge of ‘how to do it’ can be

condensed into more or less comprehensive solutions and sub-solutions that agents are

content with, but it is more comprehensive. It also includes knowledge about how not

to do it.

We are convinced that the concept of routine becomes most fundamental and powerful when

it is defined as an antipode to decision making. Only than, routine acquires the fundamental

status as a second principle of act determination and as a second pillar of microeconomic

explanation. To become an antipode, routine has to serve the same function than decision

making. However, the one and only function of decision making is to determine what

operations to conduct to bring about a given intended state. Here, the decision framework

indeed offers an alternative: replication. Substance of replication is the application of previous

solutions. Moreover, it is only useful to apply previous solutions when agents were content

with their results. However, if this is the case, replication indeed has the potential to substitute

planning and decision making. Than agents indeed have the alternative either to go through

the planning procedure anew or to apply a previous solution without further planning.

It only remains the question of how precisely attribute routine to repetition. One possibility

is certainly to use replication and routine as synonyms. However, such a definition would

exclude the possibility that have routine but refrain from utilizing it. For this reason, we

propose to approach the phenomenon a little differently and define routine as a person’s

ability to substitute deliberate planning and decision making with replication. By ability, we

do not mean the mere capacity to replicate the act in any manner and irrespective of its

results, but instead replication in a well-founded manner. The less planning is necessary to

meet our aspirations, the greater our ability to replicate the act. Likewise, the better the result

we can achieve without planning, the greater our ability to replicate the act. By replication,

we mean the application of former solutions to determine what operations to conduct to bring

about a current intended state. By substitution we do not mean immediate replacement, but

gradual substitution.

Substance of routine is specific knowledge combined with the ability to apply this

21

knowledge in specific situations. However, agents (at least principally) have the degree of

freedom to apply their routine in specific situations or not. The main consequence of the

application of routine is absence of planning. For this reason, the main result of the

acquisition of routine is to save deliberation effort so that it constitutes a very effectively as a

strategy to deal with the human bounds of rationality. And for this reason, routine has no

meaningful content in traditional decision theory. If the deliberation capacity is unrestricted

strategies to deal with human bounds of rationality are not required and incomplete planning

is irrational (since it prevents agents from processing relevant information and leads them to

sub-optimal results). However, it has to be noted that that the improvement of results and

decrease of uncertainty are not results of the acquisition of routines. They are results of

learning in the course of repetition which also takes place (perhaps to a greater extent since

more information is processed) when no routine is applied.12

It is certainly unusual and perhaps even a little irritating to define routine as an ability. But in

defense, it might be emphasized that this definition does not imply major changes in the

general understanding of the phenomenon. Most of all, the basic characteristics of routines

remain untouched by our action-based approach. In accordance with Winter (2003: 991), also

the action-based approach describes routines (in the informal sense of acts that are somewhat

dominantly controlled by replication) as ‘behaviour that is learned, highly patterned,

repetitious, or quasi-repetitious, [and] founded in part in tacit knowledge.’ The crucial

advantage of the action-based approach, however, is to produce these characteristics in act as

an outcome of the interplay of purpose, repetition, learning, deliberation and replication in the

face of the given bounds of human rationality. It therefore has the potential to explain why

and how economic agents develop patterns in act, how stable these patterns are, and how they

change in the course of repetition.

Let us finally do some words on the individual-organizational question. Some authors, in

12 It might be noted that this cognitive approach allows to define ‘a routine’ informally as an act that is

predominantly determined by replication (or highly routinized). This would not only transform routine back to

act and create an equivalent to the definition of the pattern approach. It also create a very convenient tool for the

analysis of specific settings. However, this notion is informal since it does not built on clear specification of

what ‘predominantly’ means.

22

particular Hodgeson (2004), define routine exclusively as an organizational phenomenon and

denote the individual phenomenon as habit. The most important implication of such a

terminological distinction is that individual habits and organizational routine are not basically

two forms of the same, but two fundamentally different phenomena. Additionally, these

authors usually declare the organizational level being the more fundamental one. However,

we fundamentally disagree with such a terminological setting for at least three reasons:

• First it breaks with the custom of every day’s use of the word. English language very

well uses the word routine for individuals in the sense of ‘I have a routine.’ Best

indication for that is that even the proponents of the distinction talk of ‘organizational

routines,’ which would be a pleonasm according to their own understanding of the

word.

• Second, the underlying phenomenon is universal and basically the same for both,

individuals and organizations: the substitution of decision making by replication. Any

distinction would be artificial and misrepresent the structure of the underlying

phenomena.

• Third, to understand routine on the organizational level, one has to progress to

individual intentions, bounds of rationality, knowledge and solutions. We have to

understand the individual incentives and obstacles when analyzing efforts to overcome

organizational routines and individual resistance against these efforts. Hence, we are

convinced that the individual level is the more fundamental one. Organizational

routine has to be understood as interplay of routines and decision making of different

individuals.

CONCLUSIONS: THE INTERPLAY OF ROUTINE AND DECISION MAKING IN

ACT

The most important conclusion of this paper is that indeed both, routine as well as decision

making are essential elements of action taking. Both serve the same purpose: to determine

what operations to conduct to bring about an intended state. But they serve this purpose in a

fundamentally different way. Decision making is a choice procedure in that information is

processed and matched with preferences. Routine is an ability to recall and apply solutions

and to shortcut decision making procedures in a qualified way.

They complement each other since routine does not suit to process new information. It is

only acquired and applied in the course of repetition. To master ‘first acts’ or to change

23

established routine acts purposefully, agents have to rely on decision making procedures.

However, decision making requires a lot of deliberation – deliberation that agents do not have

for every act since their rationality is bounded. In other words: it is simply not possible for

agents to base all their acts on comprehensive decision making. By shortcutting decision

making procedures in a qualified way, routine provides a necessary strategy to deal with the

given bounds of rationality. Hence without decision making there would be no purposeful

change but without routine, agents were completely overwhelmed by the sheer complexity of

reality. For this reason both, routine and decision making, matter to explain act.

However, routine and decision making are not independent, but mutually influence each

other. On one hand, the application of routine shortcuts decisions making procedures and

simplifies planning processes with every sub-solution that has been established. Established

sub-solutions are not compared with alternatives anymore, but chosen on basis of the

degenerated decision ‘do it like last time.’ In this sense, routine has the potential to substitute

decision making. The ‘routine trick’ is that this substitution is qualified since solutions

condense the results of previous planning and are only chosen when agents are content with

their results. However, necessary precondition for every acquisition and application of routine

is repetition and this requires a stable context and a stable intention.

On the other hand, decision making ‘drives’ the acquisition of routine and basically

‘controls’ its application. The decision is the basis for the operation and both together are the

basis for the acquisition of knowledge, which again is the basis for the acquisition of routine.

If agents decide for a different alternative, they necessarily acquire a different routine. For this

reason, decision making ‘drives’ the acquisition of routine. It ‘controls’ the application of

routine since agents basically always have the opportunity to break with the established

routine and re-plan acts (or take completely new acts). However, this control of act by

decision making is much less comprehensive than assumed by traditional decision theory.

REFERENCES

Anscombe, G.E.M. (1957), Intention, Oxford: Basil Blackwell.

Arrow, J.K. (1962), The economic implications of learning by doing. Review of Economic

Studies 29: 155-173.

Becker, M. (2005): The concept of routines: some clarifications, Cambridge Journal of

Economics (29): 249-262.

24

Becker, M.C. and Knudsen, T. (2004) The Role of Routines in Reducing Pervasive Uncer-

tainty, Journal of Business Research 58: 746– 757.

Betsche, T., Brinkmann, B.J., Fiedler, K., Breining, K. (1999), When prior knowledge

overrules new evidence: Adaptive use of decision strategies and the role of behavioral

routines, Swiss Journal of Psychology 58: 151-160.

Betsch, T. (2005), Wie beeinflussen Routinen das Entscheidungsverhalten?, Psychologische

Rundschau 56: 261-270.

Cohen, M., Burkhard, R., Dosi, G., Egidi, M., Marengo, L., Warglien, M., and Winter, S.G.

(1996), Routines and Other Recurring Action Patterns of Organizations: Contemporary

Research Issues, Industrial and Corporate Change 5: 653-98.

Conlisk, J. (1996), Why bounded rationality? Journal of Economic Literature 34: 669-700.

Davidson, D. (1963), Actions, Reasons, and Causes, Journal of Philosophy 66: 685-700.

del Val, M.P. and Martínez Fuentes, C. (2003), Resistance to change: a literature review and

empirical study, Management Decision 41: 148-55.

Dosi, G., Teece, D., and Winter, S.G. (1992), Toward a Theory of Corporate Coherence:

Some Preliminary Remarks, in: G. Dosi, R. Giannetti und P. A. Toninelli (eds.),

Technology and Enterprise in a Historical Perspective, Oxford: Clarendon Press, pp. 185-

211.

Dosi, G., Nelson, R.R. and Winter, S.G. (2000), The Nature and Dynamics of Organizational

Capabilities, Oxford: Oxford University Press.

Eisenhardt, K.M. and Martin, J.A. (2000), Dynamic capabilities: what are they? Strategic

Management Journal 21: 1105-1121.

Feldman, M.S. (2000), Organizational routines as a source of continuous change,

Organizational Science 6: 611-629.

Feldman, M.S. (2003), A performance perspective on stability and change in organizational

routines, Industrial and Corporate Change 12: 727-752.

Feldman, M.S. and Rafaeli, A. (2002), Organisational routines as sources of connections and

understandings, Journal of Management Studies 39: 309-331.

Hodgson, J. (2004), The Firm as an Interactor: Firms as Vehicles for Habits and Routines,

Journal of Evolutionary Economics 14: 281-307.

Jeffrey, R.C. (1983), The Logic of Decision, 2nd. Edition, Chicago and London: University of

Chicago Press.

25

Kesting, P. (2006), The Meaning of Intentionality for Decision Making, available at SSRN:

http://ssrn.com/abstract=887088.

Kuhn, T. S. 1970. The Structure of Scientific Revolutions. 2nd edition, Chicago: University

Press.

March, J.G. (1995), A Primer on Decision Theory, New York: The Free Press.

Nightingale, P. (2003), If Nelson and Winter are only half right about tacit knowledge, which

half? A Searlian critique of codification, Industrial and Corporate Change 12: 149-183.

Pentlend, B.T. (2003a), Sequential Variety in Work Processes, Organizational Science 14:

528-540.

Pentlend, B.T. (2003b), Conceptualizing and Measuring Variety in the Execution of

Organizational Work Process, Management Science 57: 857-870.

Pentland, B. T., Feldman, M. (2005), Organizational routines as a unit of analysis, Industrial

and Corporate Change 14: 793-815.

Roy, O. (2006), Commitment-Based Decision Making for Bounded Agents, working paper

pp-2006-34 of the Institute for Logic, Language & Computation, Universiteit van Amster-

dam. http://www.illc.uva.nl/Publications/ResearchReports/PP-2006-34.text.pdf.

Runde, J. (2002), Filling in the Background, Journal of Economic Methodology 9(1): 11-30.

Samuelson, P.A. (1983), Foundations of Economic Analysis, enlarged edition, Cambridge

(Mass.) and London: Harvard University Press.

Savage, L. J. (1954), The Foundations of Statistics, New York: Dover.

van Hees, M. (2003). Intentions, utility and rationality, working paper,

http://www.philos.rug.nl/~martin/IntentieJan2003MSMetCorr.pdf.

Simon, H.A., March, J.G. (1958), Organizations, Wiley: New York.

von Wright, G.H. (1971), Explanation and Understanding, London: Routledge & Kegan Paul.

Winter, S.G. (1986), The Research Program of the Behavioral Theory of the Firm: Orthodox

Critique and Evolutionary Perspective, in: Gilad, B. and Kaish, S. (Eds.) Handbook of

Behavioral Economics, Vol. A, Greenwich: JAI Press, pp. 151-188.

Winter, S.G. (2003), Understanding Dynamic Capabilities, Strategic Management Journal 24:

991-995.