CogReact: A Reinforced Framework to Model Human Cognitive Reaction Modulated by Dynamic Intervention

Songlin Xu, Xinyu Zhang

TL;DR: This paper presents a computational framework integrating cognitive models and data-driven deep reinforcement learning to model human cognitive response modulated by dynamic environmental stimuli.

Abstract: Using deep neural networks as computational models to simulate cognitive processes can provide key insights into human behavioral dynamics. Challenges arise when environments are highly dynamic, obscuring stimulus-behavior relationships. However, the majority of current research focuses on simulating human cognitive behaviors under ideal conditions, neglecting the influence of environmental disturbances. We propose CogReact, which integrates drift-diffusion with deep reinforcement learning to simulate granular effects of dynamic environmental stimuli on the human cognitive process. Quantitatively, it improves cognition modeling by considering the temporal effect of environmental stimuli on the cognitive process and captures both subject-specific and stimuli-specific behavioral differences. Qualitatively, it captures general trends in the human cognitive process under stimuli. We examine our approach under diverse environmental influences across various cognitive tasks. Overall, it demonstrates a powerful, data-driven methodology to simulate, align with, and understand the vagaries of human cognitive response in dynamic contexts.

Lay Summary:

Understanding how people make decisions in ever-changing environments is a major challenge in psychology and artificial intelligence. Most existing computer models that simulate human thinking mainly work well under ideal, controlled conditions — but real life is messy, with unpredictable events and shifting contexts.

To tackle this, we created CogReact, a new system that combines deep reinforcement learning (a kind of data-driven AI model) with classical models from cognitive science. Our approach captures how people react over time to changing surroundings and how those changes affect their cognitive processing.

CogReact doesn’t just model general trends; it also picks up on how different people and different types of situations lead to different behaviors. It outperforms existing models in both precision and realism.

We tested CogReact across a variety of thinking tasks in dynamic settings, showing it can mimic human behavior more accurately than previous methods. This means we’re one step closer to building AI systems that understand how humans think — not just in theory, but in the unpredictable environments of the real world.

Link To Code: https://github.com/songlinxu/CogReact

Primary Area: Applications->Neuroscience, Cognitive Science

Keywords: Cognitive Response, Deep Reinforcement Learning, Dynamic Stimuli

Flagged For Ethics Review: true

Submission Number: 2006