Artificial Intelligence and Agency: Tie-breaking in AI Decision-Making

Area of Science:

• Artificial Intelligence
• Philosophy of Mind
• Machine Ethics

Background:

• The increasing integration of artificial intelligence (AI) into society necessitates a clear understanding of machine agency.
• Defining agency is crucial for developing appropriate policies and narratives for human-machine coexistence.
• Decision-making processes are central to agential action, often leading to favorable outcomes.

Purpose of the Study:

• To identify hallmark features of agency, particularly in the context of decision-making.
• To investigate whether artificial intelligence (AI) exhibits these features, especially in tie-breaking scenarios.
• To determine if current AI possesses the capacity for voluntarist reasoning.

Main Methods:

• Conceptual analysis of agency and decision-making processes.
• Examination of AI's current tie-breaking mechanisms.
• Philosophical argumentation regarding voluntarist reasons as a marker of agency.

Main Results:

• Decision-making processes are fundamental to agency, generally producing sound choices.
• Agents may encounter situations with equally valid choices, resulting in a tie.
• Current AI tie-breaking methods do not demonstrate the ability to generate voluntarist reasons.

Conclusions:

• The capacity to create voluntarist reasons in tie situations is a defining characteristic of agency.
• Artificial intelligence, as currently designed, does not fulfill this criterion for agency.
• Further research is needed to bridge the gap between AI capabilities and human-like agency.