agency in human-machine networks; impacts on trust · pdf fileagency in human-machine...
TRANSCRIPT
Agency in Human-Machine Networks; Impacts on Trust and Behaviour
21 March 2017
Dr Vegard EngenIT Innovation Centre, University of Southampton, UK
HUMANE Workshop, Oxford, UK
AGENCY
• Agency forms part of the HUMANE typology– Describes the capabilities of the human and machine actors
• Agency: a much debated topic in the literature– Several theoretical models of human agency– However, debate on whether machines can exhibit agency
– Platforms, software components/agents, devices and infrastructure, sensors, robots, …
• We have revisited definitions to be useful for the purposes of analysing, understanding and designing HMNs– Needs to be practical and useful for HMN designers
© University of Southampton IT Innovation Centre 2016 3
What is agency?• According to Structuration Theory (ST):
– “the capability to make a difference” (Giddens 1984) • According to Actor Network Theory (ANT):
– “[no] difference in kind between people on the one hand, and objects on the other” (Law, 1992)
• According to Social Cognitive Theory (SCT):– “agency refers to acts done intentionally” (Bandura 2001)
• Thus, we have a conflict…– Intentionality implies agency is exclusive to humans– This is a sticking point for some regarding actor-network theory
Bandura, A.: Social Cognitive Theory: An Agentic Perspective. Annu. Rev. Psychol. 52, 1–26 (2001)Giddens, A.: The Constitution of Society: Outline of the Theory of Structuration., (1984).Law, J.: Notes on the Theory of the Actor-Network: Ordering, Strategy, and Heterogeneity. Syst. Pract. 5, 379–393 (1992)
© University of Southampton IT Innovation Centre 2016 4
Intentionality – Bandura (2001) cont.
• Intentionality encompasses the ability to choose to behave in a certain way– in particular with a future course of action in mind.
• Related to forethought– setting goals, which affects behaviour in order to achieve desired
outcomes and avoid undesired ones.• Linked to motivation
– guides the chosen actions and anticipations of future events, which “provides direction, coherence, and meaning to one’s life” (Bandura 2001)
Bandura, A.: Social Cognitive Theory: An Agentic Perspective. Annu. Rev. Psychol. 52, 1–26 (2001)
© University of Southampton IT Innovation Centre 2016 5
So, what about the machines?• Compared to humans, machines do not…
– have self-generated intention or motivation– experience trust or reliance, nor behave altruistically or irrationally of their own volition
• However, there is a need to refer to ‘machine agency’– active participants in increasingly significant roles– influencing other agents and the outcomes of the HMN itself
• Some have argued similarly in other contexts and include machine agency in one way or another, e.g.,– Actor-Network Theory (Law 1992)– Double Dance of Agency (Rose and Jones 2005)– Increasing complexity of machine algorithms (Hildebrandt, 2015)
Law, J.: Notes on the Theory of the Actor-Network: Ordering, Strategy, and Heterogeneity. Syst. Pract. 5, 379–393 (1992).Rose, J., Jones, M.: The Double Dance of Agency: A Socio-Theoretic Account of How Machines and Humans Interact. Syst. Signs Actions. 1, 19–37 (2005).Hildebrandt, M.: Smart Technologies and the End of Law: Novel Entanglements of Law and Technology. Cheltenham, UK: Edward Elgar Publishing Ltd. (2015)
© University of Southampton IT Innovation Centre 2016 6
A practical definition
“The capacity to perform activities in a particular environment in line with a set of goals/objectives that influence and shape the extent and nature of their participation” (Engen et al., 2016)
– Whether human or machine– Environment being the HMN
Three factors that can be used to quantify agency:1. the activities the agent can perform (quantity, freedom vs restriction)2. the nature of the activities (open vs closed, predictability, emergence)3. the ability to interact with other agents (exert influence)+ perceived agency for machines
Engen, V., Pickering, J.B., and Walland, P.: Machine Agency in Human-Machine Networks; Impacts and Trust Implications. HCI International (2016).
© University of Southampton IT Innovation Centre 2016 7
Machine agency
• In practical terms, our definition of machine agency reflects the degree to which machine agents maya) perform activities of a personal and creative nature (e.g., supporting
health care by personalising motivation strategies), b) influence other agents in the HMN (thus, “make a difference”), c) enable human agents to exercise proxy agency, and d) the extent to which they are perceived as having
agency by human agents (anthropomorphism).
© University of Southampton IT Innovation Centre 2016 8
Examples of human agency
© University of Southampton IT Innovation Centre 2016 9
Many activities, open nature, much interaction Few activities, closed nature, no interaction
• Informs analysis of implications for the design of the HMN, e.g.,– Trust, user adoption, participation,
motivation, collaboration, etc.– Robustness, failover, interoperability,
integrity, security, etc.• Particularly useful for HMNs that
have different states• eVACUATE – a complex network
– Emergency decision support for crowd management
– 2 states– Normal operation (monitoring)– Crisis evacuation
© University of Southampton IT Innovation Centre 2016 10
Characterising HMNs – agency
AGENCY, TRUST AND BEHAVIOUR
Definition of trust
“the willingness of a party to be vulnerable to the actions of another party based on the expectation that the other will perform a particular action important to the trustor, irrespective of the ability to monitor or control that other party” (Mayer et al., 1995, p.712)
• In short: willingness to expose oneself to vulnerability.• Involves risk assessment to decide whether to make
oneself vulnerable (Shin, 2010).
Mayer, R.C., Davis, J.H., and Schoorman F.D.: An Integrative Model of Organizational Trust. Academy of Management Review., 20(3), 709-734 (1995)Shin, D-H.: The Effects of Trust, Security and Privacy in Social Networking: A Security-Based Approach toUnderstand the Pattern of Adoption. Interacting with Computers, 22(5), 428-438 (2010)
© University of Southampton IT Innovation Centre 2016 12
Agency and trust• Trust is important for technology acceptance (Mayer et al., 1995) and
intentions to explore technology (Thatcher et al., 2011).• Machines perform more complex and autonomous tasks increasing the
agency– Opportunities for innovation (Følstad et al., 2017).– However, implications for trust relationships.
– Are humans becoming more reliant or even dependent on the machines?– Necessary to control agency for ethical reasons? How to manage accountability & responsibility?
– Trust is increasingly important to develop and maintain.
Mayer, R.C., Davis, J.H., and Schoorman F.D.: An Integrative Model of Organizational Trust. Academy of Management Review, 20(3), 709-734 (1995)Thatcher, J.B., McKnight, D.H., Baker, E.W., Arsal, R.E., and Roberts, N.H.: The Role of Trust in Postadoption IT Exploration: An Empirical Examination of Knowledge Management Systems. IEEE Transactions on Engineering Management, 58(1), 56-70 (2011)Følstad, A., Engen, V., Haugstveit, I.M., and Pickering, J.B.: Automation in Human-Machine Networks: How Increasing Machine Agency Affects Human Agency. International Conference on Man-Machine Interactions (2017) [submitted]
© University of Southampton IT Innovation Centre 2016 13
Model of key influences on behaviour in HMNs
© University of Southampton IT Innovation Centre 2016 14
• Trust model of 7 key inter-related constructs affecting the behaviour of people in HMNs– Expanding model of Thatcher et al. (2011) into the wider ecosystem (risk, regulation and agency)
• Pickering, J.B., Engen, V., and Walland, P.: The Interplay Between Human and Machine Agency. To appear in HCI International (2017)
(Computer) self-efficacy
© University of Southampton IT Innovation Centre 2016 15
• Related to human agency, and affected by machine agency.• Positively correlated with the behaviour in the HMN, but negatively correlated with perceived
risk.
Trust
© University of Southampton IT Innovation Centre 2016 16
• Positively correlated with human agency.• Mediates the effects of human agency on the behaviour in the HMN.• Negatively affected by perceived risk.
Regulation
© University of Southampton IT Innovation Centre 2016 17
• Perceived risk and machine agency negatively correlated with changes in regulation.• Regulation affects human agency.
Use of the model
1. Informing implication analysis, e.g.,– What if we are able to reduce perceived risk?
– We can expect an increase in trust.– In turn, increase in trust influences the behaviour of people in the HMN.
– Strategies for reducing perceived risk, e.g.,– Increase regulatory control.– Encourage users’ belief in their own abilities in the HMN (self-efficacy).
2. Understanding behaviour in a network– Identifying and addressing key influences on behaviour.– E.g., people stopping to use the HMN due to a break-down in trust.
– How to re-build trust implication analysis (above).
© University of Southampton IT Innovation Centre 2016 18
CONCLUSIONS AND FURTHER WORK
Conclusions and further work
• Established practical definitions of human and machine agency• Proposed a model describing the key influences on the behaviour
of people in HMNs– Extending the work of Thatcher et al. (2011) wider ecosystem– Increase our understanding of the interplay between agency and trust,
and other key constructs such as social norms and self-efficacy• Model to be evaluated
– First via an expert group (familiar with trust, H2M interactions and technology)
– Then via a quantitative survey (extending instruments from previous research) factor analysis to validate hypotheses
© University of Southampton IT Innovation Centre 2016 20
Thank you!
Contact info: – Dr Vegard Engen– IT Innovation Centre, University of Southampton– mailto: [email protected]
© University of Southampton IT Innovation Centre 2016 21
spørsmål? Въпроси?
questions?ερωτήσεις;
؟پرسشها