Anthropomorphism on Risk Perception: The Role of Trust and Domain Knowledge in Decision-Support AI

Anthropomorphic AI design can dangerously lower risk perception in users with low domain knowledge: The study found that people without financial expertise who interacted with human-like AI agents perceived financial decisions as less risky than they actually were, potentially leading to poor decision-making in vulnerable populations.

Trust mechanisms work differently depending on user expertise: While anthropomorphism increased both "cognitive trust" (belief in competence) and "affective trust" (emotional comfort) in all users, only those with high financial knowledge benefited from these trust increases. For low-knowledge users, increased cognitive trust actually led to lower risk perception—a concerning disconnect from reality.

Friendly AI design may create false confidence in non-experts: The research suggests that making AI conversational agents more approachable and human-like can backfire for users without specialized knowledge, causing them to underestimate risks rather than make more informed decisions.

Domain knowledge is critical for responsible AI deployment: The study demonstrates that the same anthropomorphic design had opposite effects depending on financial literacy, highlighting that responsible AI systems must account for user expertise levels rather than applying one-size-fits-all design approaches.

Calibrating trust, not maximizing it, should be the goal: Rather than designing AI to be as likeable and trustworthy as possible, the authors argue that responsible AI design should aim to calibrate trust appropriately—ensuring users trust AI systems exactly as much as they should given actual capabilities and risks.