The ability to identify faulty reasoning in speech is a crucial skill in any healthy democracy. Designing instructional systems that teach this skill, however, is complicated by the unique role bias plays in our ability to evaluate political arguments. Here we offer a new paradigm for accommodating and mitigating the impact of bias on the identification of informal logical fallacies in spoken arguments. A randomized controlled experiment showed both a clear effect of bias on performance, and the positive impact of a user-adaptive intervention designed to reduce the effect of bias. These results suggest that the challenges of bias in this instructional domain may be addressed by designing systems that can adapt instruction to the values of the user.