Designing Beyond Anthropomorphism: Human- and Robot-Like Error Recovery for Robot Navigation Errors

Abstract

Robots operating in shared human environments will inevitably make mistakes. These errors can negatively affect users’ attitudes toward robots, making effective recovery strategies essential. While prior work has focused on human-like repair behaviors, these may be mismatched with a robot’s capabilities and increase perceptions of its social intelligence. It remains unclear whether robot-like recovery strategies can restore trust without increasing perceived social intelligence. We conducted an online animated video study (N = 219) in which participants observed a healthcare robot making a navigation error followed by one of four recovery strategies: none, robot-like (functional), human-like (socio-emotional), or combined. Trust and perceived social intelligence (PSI) were measured before and after the error. Both trust and PSI decreased significantly following the error. Recovery strategies restored both measures, although their effects differed. Robot-like recovery restored trust more effectively than human-like recovery while avoiding increases in PSI, whereas human-like recovery elevated PSI beyond baseline levels. The combined strategy produced the strongest overall recovery. These findings suggest that robot-like recovery are an effective approach to trust repair while avoiding the increases in perceived social intelligence associated with human-like recovery behaviors.