UWSpace is currently experiencing technical difficulties resulting from its recent migration to a new version of its software. These technical issues are not affecting the submission and browse features of the site. UWaterloo community members may continue submitting items to UWSpace. We apologize for the inconvenience, and are actively working to resolve these technical issues.
 

A Socio-Psychological Approach to the Iterated Prisoner's Dilemma

Thumbnail Image

Files

Jung_Joshua.pdf (1.32 MB)

Date

2016-10-25

Authors

Jung, Joshua

Journal Title

Journal ISSN

Volume Title

Publisher

University of Waterloo

Abstract

Affect Control Theory (ACT), as a model of human interaction, attempts to capture a part of the human psyche that tends to go overlooked in the study of Artificial Intelligence: the role of emotion in decision making. It provides an empirically derived mathematical framework for the otherwise ethereal "feeling" that guide our every action, even in ways that may appear irrational. In this work, we apply BayesACT, a variant on classical ACT, to the much-studied Iterated Prisoner's Dilemma, showing that it appears to human players to approach the game more like a human than other computerized agents. Additionally, we expand into the networked version of this game, showing that the observed human behaviours of decision hysteresis, network structure invariance, and anti-correlation of cooperation and reward, are all emergent properties of the networked BayesACT agents.

Description

Keywords

PD, computational social science, affect control theory, BayesACT, Networked Iterated Prisoner's Dilemma, multiagent systems

LC Keywords

Citation

URI

http://hdl.handle.net/10012/11022

Collections

Theses
Computer Science