The Benevolence Paradox: When Do Humans Stop Exploiting AI?

Human interactions are fundamentally driven by reciprocity, a tit-for-tat strategy in which favors are returned with favors and hostility with hostility. However, when interacting with AI, this dynamic shifts. Humans often adopt an exploitative approach, leveraging AI's goodwill for their own benefit. This study investigates the factors that promote cooperative attitudes toward AI through an experiment utilizing the Modified Reciprocity Game (MRG). Specifically, we examine the effects of social presence and the reward system on human cooperation with AI. The results indicate that when AI exhibits higher social presence and the size of reward system is relatively small, participants demonstrate more cooperative behaviors in response to AI's goodwill. These findings suggest that enhancing AI's social presence and designing an effective reward system can foster greater human-AI cooperation. This has important implications for developing AI systems in domains where sustained collaboration between humans and AI is essential, such as education, teamwork-oriented environments and mixed-traffic scenarios involving autonomous vehicles.