This research tackles the cognitive biases in large language models that limit their reliability for critical decision-making applications in healthcare and other domains.

• Identifies systematic patterns of bias that lead LLMs to produce inaccurate judgments
• Proposes debiasing techniques to enhance LLM reasoning in decision-support contexts
• Demonstrates improved performance in healthcare decision-making tasks
• Addresses concerns about reliability when LLMs serve as conversational assistants

Medical Impact: By reducing cognitive biases, healthcare professionals can more safely leverage LLMs for patient care decisions, improving diagnostic accuracy and treatment recommendations while reducing errors caused by flawed AI reasoning.

Cognitive Debiasing Large Language Models for Decision-Making