Trust in Generative AI for Health Information — Learned Dependency and Trust Calibration Failures
OVERVIEWA randomized controlled experimental study examined the relationship between user trust in AI-generated health information and a construct called "learned dependency" — the degree to which users habitually defer to AI outputs without independent evaluation. The study found that users generally trust accurate AI-generated health information more than inaccurate information, suggesting baseline discriminative capacity. However, higher self-reported learned dependency was associated with weaker trust calibration, predicting greater susceptibility to incorrect AI outputs. The trust calibration failure identified in high-dependency users is clinically significant: these individuals showed reduced ability to distinguish accurate from inaccurate health information, even when both were presented. This pattern suggests that repeated AI use may itself erode the critical evaluation skills needed to safely use AI — a self-reinforcing dynamic in which dependency increases vulnerability to harm. The study tested whether text-based visual cues (TVCs), such as text highlighting, could mitigate this effect and found them insufficient. The findings highlight a design gap: current AI health information interfaces lack effective mechanisms to support critical evaluation among high-dependency users. The authors call for more sophisticated design interventions that go beyond surface-level cues to actively scaffold user reasoning and reduce overreliance. This research has direct implications for…
Subscribe to read full wiki analysis
Full wiki content — including source provenance, contradiction tracking, related pages, and analyst notes — is available to Subscriber members.
View Plans & SubscribeNew users get a 7-day trial — no credit card required.