The problem of agency in artificial intelligence

This study aims to discuss the nature and limits of agency in the context of artificial intelligence (AI). The historical, ontological and ethical contexts of the concept of human agency are examined, and how AI can adapt to or transcend these concepts are addressed. The study distinguishes the concepts of "subject", "person" and "agent" and details the principal conditions of each: while discussing qualities such as intention, consciousness, rationality, moral responsibility, language use, continuity and the ability to attribute value, it is argued that consciousness and free will are the fundamental building blocks of agency. The agency of AI is evaluated by comparing it with different types of beings (e.g. animals) and human agency models. The study analyzes how an AI's capacity to be conscious and exhibit free will can be measured by using traditional tests such as the Turing Test. Ontological differences are addressed with a comparative approach to human agency characteristics such as AI not having a physical body and not being able to access death information. Finally, technical and philosophical issues faced by modern AI systems, such as the capacity of AI to understand natural language, limitations of logical modeling, and free will in the context of moral responsibility, are discussed. While determining the criteria required for general artificial intelligence (AI) to be accepted as an agent, the study suggests that our understanding of human agency should also be redefined in this process. Thus, the potential agency of AI is examined in both an ethical and metaphysical framework and new theoretical approaches are suggested.