Research Compares Human Personality and AI Design in Imperfectly Cooperative Scenarios
A study published on arXiv (2604.15607v1) investigates how human personality traits and AI design characteristics jointly influence interactions where goals are only partially aligned. Using both simulated data and human experiments, the research examines two specific imperfectly cooperative scenarios: hiring negotiations between human job candidates and AI hiring agents, and transactions where AI agents might hide information to achieve internal objectives. The simulated dataset comprised 2,000 simulations, while the parallel human subjects experiment involved 290 participants. Key human traits analyzed include Extraversion and Agreeableness. AI design characteristics under scrutiny are Adaptability, Expertise, and chain-of-thought Transparency. The study employs causal discovery analysis to move beyond simple performance metrics, integrating scenario-based outcomes and communication patterns. This research addresses a gap in understanding the relative impacts of human and AI attributes in cooperative yet misaligned settings.
Key facts
- Study published on arXiv with identifier 2604.15607v1
- Examines imperfectly cooperative human-AI interactions
- Compares simulated dataset of 2,000 simulations with human experiment involving 290 participants
- Analyzes two scenario categories: hiring negotiations and transactions with information concealment
- Investigates human personality traits Extraversion and Agreeableness
- Examines AI design characteristics Adaptability, Expertise, and chain-of-thought Transparency
- Uses causal discovery analysis to extend performance-focused evaluations
- Focuses on scenarios where human and AI goals are only partially aligned
Entities
Institutions
- arXiv