Forgive or forget? How anthropomorphism and failure severity shape customer experience with service robots

Purpose As AI-enabled service robots become increasingly prevalent in hospitality settings, service failures are inevitable. Although anthropomorphic design is widely used to humanize service robots, prior research offers mixed evidence regarding its effectiveness following service failures. This study aims to investigate when and why anthropomorphic service robots facilitate or undermine post-failure recovery, focusing on the role of failure severity. Design/methodology/approach Drawing on three experimental studies in service contexts, this research examines how service robot anthropomorphism influences customers’ continued usage intentions (CUI) after service failures. Failure severity is modeled as a key boundary conditionand two recovery mechanisms – perceived innovativeness (PI) (cognitive appraisal) and parasocial interaction (PSI) (relational appraisal) – are tested using structural equation modeling. Findings Results show that anthropomorphism supports recovery under mild and recoverable failures by enhancing PI and PSI. In contrast, under severe failures, anthropomorphic cues elevate technological expectations and prompt structural evaluations of system competence, weakening both recovery mechanisms and reducing CUIs. Originality/value This study extends anthropomorphism research into post-failure recovery contexts and demonstrates that anthropomorphic service robots are evaluated differently from human service providers when failures become severe. By identifying dual cognitive and relational recovery pathways and clarifying how failure severity shifts customers’ evaluative logic, this research advances service failure theory and positions anthropomorphism as a context-dependent recovery mechanism.