This repository holds the data, modeling code, and writeup for our final project for COS454, Computation Models of Cognition. Authors: Colton Loftus, Maya Rozenshteyn
Cognitive models often assume that humans optimize tasks in a generally rational manner. Analyzing human task-related behavior under a state of fatigue, however, may challenge these assumptions of rationality. Physical and mental fatigue may alter not only task performance but also fundamental approaches to task-solving. This paper explores behavior during the two-stage Markov decision task completed under conditions of differing fatigue. Analysis of fatigue-modulated behavior in this task illustrates statistically significant differences in people's use of model-based versus model-free reinforcement learning strategies depending on fatigue level. Specifically, decreased fatigue results in decreased usage of both model-based and model-free strategies.
All references and related work for this project can be sound in our bibliography Our two-step task code is a modified version of the implementation by the Princeton University Nivlab (Author(s): Sam Zorowitz, Gili Karni, Branson Byers) found here