Finding: Bayesian modeling shows that unanimous group agreement is *not* the most reliable signal of group accuracy β groups with high but not total consensus are most trustworthy, because conformity bias inflates false unanimity.
Why it matters: Directly relevant to team dynamics and collective intelligence: challenges the intuition that consensus = correctness, with implications for how teams make decisions, how experts communicate recommendations, and how conformity bias degrades group epistemic quality.
Method: Theoretical Bayesian modeling incorporating conformity bias as a parameter across groups with varying agreement levels.
Jargon: *Conformity bias* β tendency to align one's expressed judgment with the group majority, reducing information diversity; *Bayesian conformist decision-maker* β an agent who rationally updates on both private evidence and social signals, but with a bias toward agreeing with peers.
Finding: Argues that education systems must balance AI integration with preserving human elements, framing the challenge as developing AI-managing rather than AI-dependent learners.
Why it matters: Touches on AI's impact on learning and human agency, but the framing is largely polemical/policy-oriented rather than empirical or theoretically substantive for HCI, behavioral science, or learning sciences research.