Humans learn to prefer trustworthy AI over human partners

AI transparency enables human learning: When bot identity was disclosed, humans initially avoided AI partners but gradually learned to prefer them as they observed consistent trustworthy behavior over time—suggesting that transparency about AI presence, while creating initial skepticism, ultimately supports informed decision-making.

Humans struggle to identify AI without disclosure: Participants misattributed bot behavior to humans and human behavior to bots when identities were hidden, indicating that AI systems can be deceptively similar to humans in communication style, which raises concerns about the need for clear AI labeling in real-world interactions.

Trustworthiness matters more than origin: Bots powered by large language models demonstrated more prosocial behavior than human partners, and humans eventually selected them based on this demonstrated reliability rather than preferring humans by default—highlighting that AI systems' actual conduct, not their nature, drives human trust.

AI competition changes human behavior: The presence of AI partners competing for selection appears to influence how humans behave in cooperative settings, suggesting that hybrid human-AI systems may reshape social norms and interaction patterns in ways that warrant careful monitoring.

Design choices significantly impact AI-human dynamics: The decision to disclose or conceal AI identity fundamentally altered partnership outcomes and learning patterns, underscoring that responsible AI deployment requires intentional choices about transparency and how humans are informed about their interaction partners.