Avoiding excessive AI service agent anthropomorphism: examining its role in delivering bad news

Abstract

Purpose

The aim of this paper is twofold. First, it seeks to understand how different forms of anthropomorphism, namely verbal and visual, can enhance or detract from the subjective well-being of consumers and their co-creation behaviors whilst collaborating with artificial intelligence (AI) service agents. Second, it seeks to understand if AI anxiety and trust in message, function as primary and secondary consumer appraisals of collaborating with AI service agents.

Design/methodology/approach

A conceptual model is developed using the theories of the uncanny valley and cognitive appraisal theory (CAT) with three hypotheses identified to guide the experimental work. The hypotheses are tested across three experimental studies which manipulate the level of anthropomorphism of AI.

Findings

Results demonstrate that verbal and visual anthropomorphism can assist consumer well-being and likelihood of co-creation. Further, this relationship is explained by the mediators of anxiety and trust.

Originality/value

The empirical results and theorizing suggest verbal anthropomorphism should be present (absent) and paired with low (high) visual anthropomorphism, which supports the “uncanny valley” effect. A moderated mediation relationship is established, which confirms AI anxiety and trust in a message as mediators of the AI service agent anthropomorphism-consumer subjective well-being/co-creation relationship. This supports the theorizing of the conceptual model based on the “uncanny valley” and CAT.