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
