Debias It Yourself: Teaching LLMs Cognitive Bias Mitigation Interventions

2026-09-30

AI Summary

This study addresses the inherent trade-off between mitigating cognitive biases and preserving reasoning capabilities in large language models (LLMs) by proposing DIY, a novel debiasing framework. For the first time, this work systematically transfers five human cognitive interventions validated in social psychology to LLMs, establishing a cognitive grounding mechanism. These interventions are implemented through three paradigms: demonstration via in-context exemplars, training through instruction tuning, and correction using guided self-reflection. Experimental results demonstrate that the proposed method achieves state-of-the-art performance across multiple benchmarks, reducing average bias to merely 2% and decreasing bias on unseen dimensions by 14.8%, while maintaining 90% reasoning accuracy. Consequently, this framework effectively optimizes the balance between debiasing efficacy and reasoning proficiency.

Abstract

Bias has long been studied in social psychology and cognitive science, where decades of research have produced a body of validated interventions that reduce stereotypical thinking and prejudiced responses in humans. We propose Debias It Yourself (DIY), a cognitively grounded framework that translates five such interventions into debiasing procedures for large language models and delivers them through three established paradigms: Show (in-context examples), Train (instruction tuning), and Revise (guided self-revision). Across three models, five bias benchmarks, eleven debiasing baselines, and three reasoning benchmarks, Train+Revise and Revise alone attain the top two average ranks, lead the bias-reasoning tradeoff (mean bias as low as 2% at 90% reasoning accuracy), and reduce bias on unseen dimensions by up to 14.8%. Our code and data are publicly available.

Problem

• Large Language Models
• Cognitive Bias Mitigation
• Debiasing
• Stereotypical Thinking
• Bias-Reasoning Tradeoff

Innovation

• Cognitive Bias Mitigation
• Large Language Models
• Instruction Tuning
• Guided Self-Revision
• Debiasing Framework