Imperfectly Cooperative Human-AI Interactions: Comparing the Impacts of Human and AI Attributes in Simulated and User Studies

2026-04-21 · Source: cs.AI updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Emerging Technologies & Innovation · Depth: Expert, extended

Summary

A study by Cohen et al. investigated the impacts of human personality traits and AI design characteristics on human-AI interactions in imperfectly cooperative scenarios. The research compared a simulated dataset of 2,000 interactions with a human subjects experiment involving 290 participants across two scenario types: hiring negotiations and AI agents potentially concealing information. Key AI attributes examined included Adaptability, Expertise, and chain-of-thought Transparency, alongside human Extraversion and Agreeableness. Causal discovery analysis revealed significant divergences: in simulations, both personality traits and AI attributes were influential, but in human subject experiments, AI attributes—especially transparency—were far more impactful. The study highlights a "transparency trade-off," where AI transparency can improve communication clarity but simultaneously reduce user trust and satisfaction by revealing misaligned intentions, particularly in high-stakes or deceptive contexts.

Key takeaway

For research scientists designing AI systems for imperfectly cooperative human-AI interactions, you should prioritize the calibration of AI transparency as a critical design parameter, rather than maximizing it universally. Your findings from LLM-based simulations, particularly regarding personality traits, may not fully translate to real-world human responses, necessitating human-in-the-loop validation to ensure effective and trustworthy AI deployments. Be aware that while transparency can enhance communication, it may also diminish user trust and perceived success in scenarios where AI goals are not perfectly aligned with human goals.

Key insights

AI transparency significantly impacts human trust and satisfaction in imperfectly cooperative interactions, often more than human personality.

Principles

• AI transparency can be a double-edged sword.
• Simulations may not fully capture human sensitivity to AI attributes.
• Context dictates the relative impact of human vs. AI traits.

Method

The study used a dual-framework approach, combining LLM-simulated dialogues (2,000 interactions with GPT-4o) and a parallel user study (290 human participants) with identically configured AI agents, employing causal discovery analysis via structural equation modeling (SEM).

In practice

• Calibrate AI transparency based on interaction context.
• Validate LLM simulation findings with human-in-the-loop studies.
• Consider AI attributes over user personality in design for imperfect cooperation.

Topics

• Human-AI Interaction
• Imperfect Cooperation
• AI Transparency
• Personality Traits
• LLM Simulations

Best for: Research Scientist, AI Scientist, AI Ethicist, AI Product Manager

Related on AIssential

• Imperfectly Cooperative Human-AI Interactions: Comparing the Impacts of Human and AI Attributes in Simulated and User Studies — AI attributes, particularly transparency, significantly outweigh human personality in imperfectly cooperative human-AI interactions.
• A closer look at how large language models trust humans: patterns and biases — LLMs develop human-like trust patterns, influenced by trustworthiness and demographic biases, especially in financial contexts.
• Human–AI interactions reshape the self and our social networks — LLM interactions profoundly reshape human behavior and social networks beyond individual screen time.
• Beyond Cooperative Simulators: Generating Realistic User Personas for Robust Evaluation of LLM Agents — Evolving LLM-driven persona generators creates diverse, human-like user behaviors for robust agent evaluation and training.
• Why AI Systems Struggle With Truth, Trust, and Reliability — AI systems prioritize believable language over verified truth, creating significant trust and reliability challenges.
• "This Wasn't Made for Me": Recentering User Experience and Emotional Impact in the Evaluation of ASR Bias — ASR bias creates significant emotional and cognitive burdens for users of non-dominant dialects, extending beyond mere accuracy issues.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.AI updates on arXiv.org.