Effects of Interoceptive Inference in Mixed-Motive Human-Robot Teams

Humans are biased against cooperating with robots and machines under mix-motives. Current interventions rely heavily on exteroceptive cues from robots to provide social cues to human teammates. While exteroceptive cues are known to be effective, no prior evidence describes how AI or robot teammates can support interoceptive inference in humans. This study aimed to understand the role of interoceptive inference with robot teammates. Findings suggested that symbolic representations of internal states may be insufficient to reduce exploitation bias toward artificial agents. More broadly, the findings differentiate framing from embodied interoceptive social cues.