When Human-Like Chatbots Backfire: Analysis of Anthropomorphism’s Dual Effects on Human Escalation in Indian Banking Services

This study examines how anthropomorphic chatbot design shapes customers' decisions to continue automated banking interactions or escalate to human agents in Indian banking services. Drawing on Privacy Calculus Theory and Social Response Theory, the study proposes that anthropomorphism has dual effects: it increases perceived competence, but also heightens perceived privacy risk. A quantitative survey of banking chatbot users was analyzed using structural equation modeling and logistic regression. The results show that anthropomorphism significantly enhances perceived competence and privacy risk. Perceived competence reduces escalation, and privacy risk does so as well in this context, indicating that privacy concerns do not necessarily trigger a switch to human agents. Time pressure strengthens the competence pathway, suggesting that customers under time pressure rely more heavily on anthropomorphic cues when judging a chatbot's capability. Task criticality weakens the influence of privacy risk on escalation, showing that high-stakes service situations shift attention toward resolution rather than risk avoidance. Overall, the findings demonstrate that anthropomorphic design produces context-dependent, not uniform, outcomes. The study contributes to theory by integrating competence and privacy mechanisms. It offers practical guidance for designing AI interfaces in banking that balance efficiency, trust, and data-related concerns for managers and policymakers in emerging market digital service environments.