MSDM Workshop @ AAMAS-09 Two Level Recursive Reasoning by Humans Playing Sequential Fixed-Sum Games Authors: Adam Goodie, Prashant Doshi, Diana Young Depts. of Psychology and Computer Science University of Georgia
Outline Introduction Experimental study Results Discussion Recursive reasoning Related work Experimental study Problem setting Participants Methodology Results Discussion
Recursive reasoning Strategic recursive reasoning in multi-agent settings (what do I think that you think that I think...) Multi-agent decision making frameworks RMM I-POMDP Theory-of-Mind Real-world application settings UAV
Related work (I) Harsanyi (1967) Mertens and Zamir (1985) agent types common knowledge Mertens and Zamir (1985) hierarchical belief system Aumann (1999) recursive beliefs
Related work (II) TOM and Behavioral game theory Stahl and Wilson (1995) a symmetric 3×3 matrix game 4% of subjects attributed recursive reasoning to their opponents Hedden and Zhang (2002) a sequential, two player, general-sum game(Centipede game) subjects predominantly began with first-level reasoning low percentage of subjects use second-level reasoning, when pitted against first-level co-players Ficici and Pfeffer (2008) a 3-player, oneshot negotiation game subjects reasoned about others while negotiating insufficient evidence to distinguish whether level two models better fit the observed data than level one models
Experimental study Problem setting Participants Methodology Opponent models Payoff structures Design of task
Problem Setting Two-player alternating-move Fixed-sum Complete and perfect information
UAV cover story scenario
Probabilities for players I and II in the cover story scenario
Participants 162 subjects Undergraduate students enrolled in lower-level Psychology courses at the University of Georgia Incentives performance-contingent monetary rewards partial course credit
Methodology Opponent models Payoff structure Design of task myopic predictive Payoff structure Design of task training phase test phase
Opponent models Myopic (First-level reasoning) Player II chooses its action based on the outcomes at states B and C
2. predictive (Second-level reasoning) Player II chooses its action by reasoning what player I will do rationally.
Payoff Structure trivial games diagnostic game D < C < B < A A < B < C < D diagnostic game C < B < A < D Different action choices for different opponent models
Design of Task Training phase trivial games criterion initial phase no rationality errors in the 5 most recent games initial phase 15 games kickoff failed to meet the criterion after 40 total training games
Test phase 40 diagnostic games intersperse with 40 C < A < B < D and D < B < A < C groups based on opponents half against myopic ones half against predictive ones In each opponent model group half played with abstract version half played with the UAV cover story and the abstract version
Results Time period Monetary incentives Training Phase Test Phase three months(September-November 2008) Monetary incentives 50 cents/correct action, average $30/participant Training Phase 162 subjects ( 26 kicked off) Test Phase 136 participants (70 female)
More accurate choices when opponent is predictive model No significant difference for two versions of games mean proportion of accurate choices across all participants in each of the 4 groups
mean proportions marginalized over the abstract and realistic versions
mean proportions marginalized over myopic and predictive opponents
Count of participants grouped according to different proportions of accurate choice
Discussion UAV cover story neither improved nor reduced the performance This particular cover story had no effect Did the subjects employ Minimax or Backward Induction? Exit questionnaire revealed most subjects did not use these Independent evaluators concurred that most subjects thought recursively In some settings humans tend to reason at higher levels of recursion
Thank you Questions?