Strategic Companions: Behavioral Evidence of Trust and Reciprocity in Human-Robot Economic Games

Abstract

We study trust and reciprocity in human-robot interaction using behavioral economic games. In Study $1 (\mathrm{N}=42)$, participants played a three-round trust game both with a human and a LuxAI QT robot, as sender or receiver. Senders invested comparable amounts in human and robot partners, but receivers returned significantly less to robots than humans across rounds, indicating attenuated reciprocity toward robots over time. MDMT scores showed moderate trust in the robot; open-ended comments suggested weaker moral accountability attributed to robots. In Study $2 (\mathbf{N} \boldsymbol{=} \mathbf{4 2})$, we tested whether a robot's pre-game favor (bonus points) elicits reciprocity. The favor reduced rather than increased transfers; return decisions largely mirrored the robot's own behavior. These findings indicate that while initial trust extends to robots, reciprocal responses are more fragile and context-dependent.