APPRAISAL-GUIDED PROXIMAL POLICY OPTIMIZATION: MODELING PSYCHOLOGICAL DISORDERS IN DYNAMIC GRID WORLD

Abstract

The increasing integration of artificial intelligence (AI) across multiple domains has high lighted the significance of comprehending and replicating human-like cognitive processes in AI. The incorporation of emotional intelligence into AI agents enables the evaluation of their emotional stability, thereby enhancing their resilience and dependability in critical decision making tasks. This work aims to develop a methodology for modeling psychological disorders using Reinforcement learning agents. Appraisal theory has been utilized to model cognitive appraisals of RL agents and train them in a dynamic grid world environment by develop ing an appraisal-guided Proximal Policy Optimization (PPO) algorithm. Further, numerous reward-shaping strategies to regulate the behavior of agents and hence simulate psycholog ical disorders have been investigated. An in-depth comparison of various configurations of the modified PPO algorithm is carried out to identify variants that can simulate Anxiety disorder and Obsessive Compulsive Disorder (OCD) like behavior in agents. In addition, an analysis of the behavioral patterns of the agents in a series of complex test environments is conducted, to evaluate the symptoms of disorders. Consequently, an effort has been made to develop a variety of evaluation criteria and metrics for analyzing the behavior of agents. Fi nally, the future possibilities and scope of studying and analyzing the psychology of artificial agents within the contexts of AI and psychology are discussed