Large Language Models (LLMs), through AI agents, represent a great opportunity for modeling the priors that negatively impact humans’ critical thinking and use this model to optimize the human-AI interaction loop. In this research project, we build upon the research on AI-enhanced reasoning by modeling a specific user’s prior beliefs, facts, and behavioral trends and deduce how their personal heuristics negatively impact their critical thinking and decision-making. We aim to (1) test whether the AI system can accurately deduce information about these priors, (2) test if the AI system can accurately detect moments where the user might be vulnerable to these biases, and (3) test if awareness of these biases significantly impacts user decision-making processes.

This SuperUROP is an occasion for me to keep doing research in AI. My previous research experience both in UROPs and internships confirmed that I enjoyed the complexity and intricacies of research. I want to keep improving as a researcher, as I plan to pursue a career in AI research. I hope that my work this year will improve the understanding of the human-AI relationship, and create more awareness about how AI can enhance the human experience.