# Are you willing to forgive generative AI doctors? Trust repair after failures in online health consultation services

## TL;DR

This paper explores how people can be encouraged to trust AI doctors again after service failures by examining factors like attribution style, social support, and anthropomorphism.

## Contribution

The study introduces a novel framework for trust repair in generative AI doctors through experimental manipulation of attribution, support, and anthropomorphism.

## Key findings

- Internal attribution, emotional support, and anthropomorphism increase trust repair in AI doctors.
- The effectiveness of social support depends on attribution style: informational support works best with internal attribution, emotional support with external attribution.
- Trust repair fully mediates the effect of social support on behavioral intention toward AI doctors.

## Abstract

While generative AI doctors are increasingly used in online health consultation services, research on trust repair following service failures remains limited. We examined how attribution style, social support, and anthropomorphism influence individuals’ trust repair and behavioral intention. A total of 512 participants were recruited to take part in a between-subjects experiment with a 2 (internal vs. external attribution) × 2 (informational vs. emotional support) × 2 (anthropomorphism vs. non-anthropomorphism) design. The results revealed that participants exposed to internal attribution, emotional support, or anthropomorphism conditions reported higher levels of trust repair. Anthropomorphism influences the effectiveness of attribution style and social support in repairing trust in GAI doctors. Moreover, an interesting interaction was observed between attribution style and social support: when the GAI doctor used internal attribution, informational support was more effective; under external attribution, emotional support proved more effective. In addition, the effect of social support on behavioral intention was fully mediated by trust repair. These findings offer practical implications for optimizing the design of GAI doctors, enhancing communication and collaboration between GAI doctors and users, and ultimately strengthening the resilience of AI-based health consultation services.