Friendly chatbot apologies backfire with shoppers in four experiments

The authors trace the effect to perceived fairness: customers judged the friendly recovery as less fair when exchange norms stood out. Anthropomorphism, attribution of responsibility and perceived control did not explain the effect.

The same warm cues can have the opposite effect depending on the relationship norms in play. For companies that deploy chatbots in e-commerce, that means recovery tone needs to match the exchange context. The findings come from exchange-oriented e-commerce failure scenarios.

The research tests a norm-contingent account of chatbot service recovery. It holds that a warm, communal chatbot framing is judged against the dominant relationship logic of the exchange, and that a mismatch makes the recovery feel unfair.

The four experiments progressed in stages: an existence proof of the backfire, isolation of the mechanism, diagnosis of the boundary conditions, and a high-fidelity behavioural replication. Across them, perceived fairness mediated the drop in forgiveness, repurchase intention and acceptance.

Because norm salience reverses how identical recovery cues are read, the authors offer a falsifiable, design-relevant account of how chatbot recovery should be governed in AI-mediated digital commerce.