Inequity aversion toward AI counterparts.

Area of Science:

• Psychology
• Human-Computer Interaction
• Artificial Intelligence Ethics

Background:

• Human moral interactions typically involve equitable resource allocation.
• The application of these moral assumptions to artificial intelligence (AI) entities is not well understood.
• Investigating human responses to AI fairness is crucial for developing ethical AI.

Purpose of the Study:

• To examine how humans respond to fair, disadvantageous, and advantageous offers from AI compared to human counterparts.
• To analyze behavioral, physiological, and affective responses in human-AI interactions.
• To propose a model for understanding human responses to AI moral behavior.

Main Methods:

• Utilized a 21-round Ultimatum Game to simulate resource allocation scenarios.
• Compared participant responses to offers from an AI versus a human counterpart.
• Measured behavioral (rejection rates), physiological (heart rate variability), and affective (emotional responses) data.

Main Results:

• Participants rejected disadvantageous offers from AI more frequently than from humans.
• Participants rejected advantageous offers from humans more frequently than from AI.
• Negative affect and physiological stress responses (heart rate variability) were heightened for disadvantageous AI offers.

Conclusions:

• Human responses to AI fairness differ significantly from responses to human fairness.
• Self-regulatory processes play a key role in how humans perceive and react to AI moral behavior.
• Findings inform the development of more ethically aligned AI systems.