Stingy bots can improve human welfare in experimental sharing networks

Area of Science:

• Social network analysis
• Artificial intelligence ethics
• Behavioral economics

Background:

• Artificial intelligence (AI) is increasingly integrated into social networks, managing resources.
• AI's resource allocation may not benefit human welfare if network dynamics are ignored.

Purpose of the Study:

• To investigate how different artificial agent (bot) allocation strategies impact human welfare in social networks.
• To determine the effect of bot network position on resource distribution and collective outcomes.

Main Methods:

• An online experiment with 496 participants in 120 human networks playing a resource-sharing game.
• Introduction of AI agents (bots) with two distinct allocation policies: reciprocal (sharing all resources) and stingy (sharing no resources).
• Manipulation of bot network positions within the human groups.

Main Results:

• Reciprocal bots had minimal impact on unequal resource distribution among humans.
• Stingy bots, when strategically positioned, balanced structural power and enhanced collective welfare in human groups.
• Stingy bots did not directly transfer wealth to human participants.

Conclusions:

• Machine allocation behavior in social networks must consider human reciprocity and interdependence.
• The effectiveness of AI in promoting human welfare is contingent on network structure and AI design.
• Thoughtful AI integration is crucial for beneficial outcomes in human-AI resource-sharing networks.