Technical Program

Paper Detail

Paper: PS-1A.37
Session: Poster Session 1A
Location: Symphony/Overture
Session Time: Thursday, September 6, 16:30 - 18:30
Presentation Time:Thursday, September 6, 16:30 - 18:30
Presentation: Poster
Paper Title: Learning and Reasoning in a Complex Environment
Manuscript:  Click here to view manuscript
Authors: David Barack, C Daniel Salzman, Columbia University, United States
Abstract: Whether negotiating social spaces, solving puzzles, or planning research, humans reason and learn in complex environments with many actions and states. From learning social hierarchies to planning a foraging route, nonhuman primates face similar problem spaces with many actions and states. Here, we report on monkeys playing a simplified version of the game Battleship, designed to investigate learning these complex environments. We focus on three questions: were monkeys able to perform this task?; what were monkeys learning during this task (e.g., shapes or movement sequences)?; and what computations were monkeys using to learn? First, monkeys were adept learners, as assessed against the optimal choice that maximized expected reward at each choice in a trial. Second, monkeys used different patterns to reveal the shape across learning, indicating that the shapes, and not movement sequences, were learned. Third, while initial model fits to the behavior show that starting information bonus models fit better than basic e-greedy reinforcement learning or gradient search models, simulated agents using these choice strategies were unable to learn the shapes (average number of choices to learn > 20,000 trials). We predict that further behavioral modelling will show that shape information as a function of entropy will reveal further aspects of monkeys' learning.