Failing to Falsify: Evaluating and Mitigating Confirmation Bias in Language Models

Ayush Rajesh Jhaveri, Anthony GX-Chen, Ilia Sucholutsky +1

Confirmation bias, the tendency to seek evidence that supports rather than challenges one's belief, hinders one's reasoning ability. We examine whether large language models (LLMs)…