Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

Agent-based modelling, laboratory experiments, and observation in the wild

19 views

Published on

An invited talk at the workshop on "Social complexity and laboratory experiments – testing assumptions and predictions of social simulation models with experiments" at Social Simulation 2018, Stockholm

Published in: Science
  • Be the first to comment

  • Be the first to like this

Agent-based modelling, laboratory experiments, and observation in the wild

  1. 1. Agent-based Modelling, laboratory experiments and observation in the wild, Bruce Edmonds, Workshop on ABM and experiments, Stockholm, August 2018 Agent-based modelling, laboratory experiments, and observation in the wild Bruce Edmonds Centre for Policy Modelling Manchester Metropolitan University
  2. 2. Agent-based Modelling, laboratory experiments and observation in the wild, Bruce Edmonds, Workshop on ABM and experiments, Stockholm, August 2018 Common-sense understanding Intuitive understanding expressed in normal language Observations of the natural system of concern Common-SenseComparison
  3. 3. Agent-based Modelling, laboratory experiments and observation in the wild, Bruce Edmonds, Workshop on ABM and experiments, Stockholm, August 2018 Scientific Understanding is Staged Intuitive understanding expressed in normal language Observations of the natural system of concern Data obtained by measuring the system Models of the processes in the system ScientificComparisons
  4. 4. Agent-based Modelling, laboratory experiments and observation in the wild, Bruce Edmonds, Workshop on ABM and experiments, Stockholm, August 2018 Experiments artificially constrained cases of the natural system Empirical experiments are another stage Intuitive understanding expressed in normal language Observations of the natural system of concern Data obtained by measuring the system Models of the processes in the system ScientificComparisons
  5. 5. Agent-based Modelling, laboratory experiments and observation in the wild, Bruce Edmonds, Workshop on ABM and experiments, Stockholm, August 2018 Models as Analogies Intuitive understanding expressed in normal language Observations of the natural system of concern Models of the processes in the system Common-SenseComparison
  6. 6. Agent-based Modelling, laboratory experiments and observation in the wild, Bruce Edmonds, Workshop on ABM and experiments, Stockholm, August 2018 Experiments as Analogies Intuitive understanding expressed in normal language Observations of the natural system of concern Common-SenseComparison Data obtained by measuring the system Models of the processes in the system Experiments artificially constrained cases of the natural system
  7. 7. Agent-based Modelling, laboratory experiments and observation in the wild, Bruce Edmonds, Workshop on ABM and experiments, Stockholm, August 2018 An Example: The Minimal Group Paradigm (MGP) Tajfel, H. (1970). Experiments in intergroup discrimination. Scientific American. 223: 96–102. A whole series of experiments where: • Subjects arbitrarily divided into groups based on some task • Given a one-shot task with tightly controlled (or no) communication or knowledge of others • Resulted in a bias in favour of their ‘own’ group • Many variations of this have been done, showing a reliable effect
  8. 8. Agent-based Modelling, laboratory experiments and observation in the wild, Bruce Edmonds, Workshop on ABM and experiments, Stockholm, August 2018 Competing Explanations Many competing explanations, including: 1. Social Identity Theory: that being part of a group was part of identity and thus important to confirm 2. Expected (but indirect) Reciprocity: there is a perception that others in a group will give back to oneself if one contributes to the group Experiments were designed by limiting options/information to distinguish between these
  9. 9. Agent-based Modelling, laboratory experiments and observation in the wild, Bruce Edmonds, Workshop on ABM and experiments, Stockholm, August 2018 Many, many experiments followed! Mostly with the motivation of finding evidence for one of the competing explanations, e.g.: • So that participants were paid a fixed sum by the experimenter vs. dependent on the game • Telling participants that they would receive the money they were awarded by in-group members only/out-group members only etc. • Participants were given ‘feedback’ that in previous trials they were favored by out-group rather than in-group members or vice versa
  10. 10. Agent-based Modelling, laboratory experiments and observation in the wild, Bruce Edmonds, Workshop on ABM and experiments, Stockholm, August 2018 Some limitations on the effect… • Oddly, in one case, when the method of contacting subjects was improved so they all arrived at the same time, the in-group bias was greatly diminished (or disappeared) • Sometimes when the game was couched specifically in terms of “payment” or “reward” the effect diminished or disappeared • There were differences between cultures as to the conditions for the expression of in-group bias • There were persistent differences between males and females (the later tending to fairness)
  11. 11. Agent-based Modelling, laboratory experiments and observation in the wild, Bruce Edmonds, Workshop on ABM and experiments, Stockholm, August 2018 Some complexities from normal life The different ‘theories’ are dynamically intertwined: • Perceived (or inferred) interdependence can give rise to categorisation as a group • Categorisation as a group can give rise to the perception (or assumption) of interdependence Thus people know, from experience, that in-groups tend to help each other, and that if you want people to help each other you form a group Fairness norms can mean in-group favouritism can be suppressed if there is no ‘excuse’ otherwise
  12. 12. Agent-based Modelling, laboratory experiments and observation in the wild, Bruce Edmonds, Workshop on ABM and experiments, Stockholm, August 2018 Some Methodological Morals • The MGP series of experiments did allow for detailed thought about what might cause what • However, the constraints imposed to decide which theory was right eliminated the natural complex dynamics which usually drove the phenomena • Many social phenomena involve similar complex temporal loops, e.g.: – Social norms (beliefs <–> observed conventions) – Status (success <–> social influence) • Reducing/limiting these can stop the natural complexes forming as they would ‘in the wild’
  13. 13. Agent-based Modelling, laboratory experiments and observation in the wild, Bruce Edmonds, Workshop on ABM and experiments, Stockholm, August 2018 Some better experiments Durrheim K, Quayle M, Tredoux CG, Titlestad K, Tooke L (2016) Investigating the Evolution of Ingroup Favoritism Using a Minimal Group Interaction Paradigm: The Effects of Inter- and Intragroup Interdependence. PLoS ONE 11 (11): e0165974. Allowed for a sequence of interactions with a known network of others seeing how in-group bias changed with time Janssen, M. A., Holahan, R., Lee, A., & Ostrom, E. (2010). Lab experiments for the study of social-ecological systems. Science, 328(5978), 613-617. Is a complex game of 4 players moving pacmen around some land and consuming ‘grass’, some treatments allowed discussion and recorded this
  14. 14. Agent-based Modelling, laboratory experiments and observation in the wild, Bruce Edmonds, Workshop on ABM and experiments, Stockholm, August 2018 A Specific Role for ABM with experiments • Because ABM can represent more complex social mechanisms – particularly those that involved dynamic and emergent patterns • This can allow for more naturalistic experiments to be done, ones that may preserve the very behaviours we are trying to understand • This implies a different experimental methodology than what is maybe usual in social psychology… • ..away from minimal settings that are easy to do, • …towards those that capture rich data and detail • …where both cognitive and social forces combine
  15. 15. Agent-based Modelling, laboratory experiments and observation in the wild, Bruce Edmonds, Workshop on ABM and experiments, Stockholm, August 2018 Conclusions • Experiments can add scientific understanding by further staging our abstraction from observations • In particular, they may be more feasible targets for social simulation than natural social phenomena • However, if the experimental conditions are too far from natural situations, they can distort or destroy the behaviour that we want to understand • Even if the setting involves real people in a real situation their behaviour can still be an artifact! • A reliable mapping to the observed world is still
  16. 16. Agent-based Modelling, laboratory experiments and observation in the wild, Bruce Edmonds, Workshop on ABM and experiments, Stockholm, August 2018 For Experiments to be Empirical Intuitive understanding expressed in normal language Observations of the natural system of concern Common-SenseComparison Data obtained by measuring the system Models of the processes in the system Experiments artificially constrained cases of the natural system Necessary NotSufficient
  17. 17. Agent-based Modelling, laboratory experiments and observation in the wild, Bruce Edmonds, Workshop on ABM and experiments, Stockholm, August 2018 The End! These slides are at: http://slideshare.net/BruceEdmonds Bruce Edmonds: http://bruce.edmonds.name Centre for Policy Modelling: http://cfpm.org

×