Managing Cognitive Biases in LLMs Utilization

Abstract

This thesis investigates the challenges posed by cognitive biases in the utilization of Large Language Models (LLMs) within the pharmaceutical industry and proposes strategies to mitigate these biases using educational tools aligned with the "Anxious" aspect of the BANI framework. Cognitive biases such as confirmation bias, automation bias, loss aversion, and fear of missing out can significantly impact decision-making, document trustworthiness, and productivity when employees rely on LLMs for generating documents.