Debiasing Politically Motivated Reasoning with Value-Adaptive Instruction

While there is a substantial appetite in the United States for improving media consumption skills, little work has focused on the biases that can make inaccurate or misleading claims feel true. This skill is particularly difficult to teach, as effective instruction requires the instructor to adapt course content to the specific beliefs of individual students, a process that is unscalable in most classrooms. Here we examine the impact of a novel method of user-centered personalized instruction that uses value-adaptivity to highlight and address user bias in the context of a civics education game. This intervention uses estimates of player and content values to predict when players may be most susceptible to biased reasoning and then intervene in those instances. We found that the intervention successfully reduced bias among high bias-regulators with practice. These results suggest that value-adaptive systems may be able to support debiasing instruction in an effective, scalable way.