Biased Information Distorts Beliefs

We often form beliefs based on information provided by others, but these sources can be biased, distorting our beliefs and decisions. Here, we examine whether individuals can detect biased sources and mitigate their influence. Participants completed a decision-making task where outcome feedback was provided by either unbiased or biased sources. Using a reinforcement learning framework, we show that participants successfully detected source biases and used this knowledge to correct biased feedback. However, these corrections were incomplete, allowing residual biases to continue shaping beliefs and decisions. Moreover, following exposure to biased sources, participants systematically misperceived unbiased sources as biased. We demonstrate how these effects can arise from a “dual-learning” process in which both beliefs about source-biases and value-estimates of choice options are updated to minimize discrepancies between outcome feedback and expectations. Our findings highlight the challenge of forming accurate beliefs and offer mechanistic insights regarding how biased environments bias our beliefs.