Competing Biases underlie Overconfidence and Underconfidence in LLMs

Nature Machine Intelligence, Published online: 22 April 2026; doi:10.1038/s42256-026-01217-9 Kumaran show that large language model (LLM) confidence is shaped by two competing biases: a choice-ive bias that inflates confidence in initial answers, and a systematic overweighting of contradictory advice, deviating from optimal Bayesian reasoning.