The appearance of free will in artificial intelligence: How human decision-making is influenced

Abstract

This thesis examines how artificial intelligence (AI), in particular large language models (LLMs), influences human deliberation and the experience of free will (FW). While much of the existing literature focuses on whether AI systems genuinely possess FW or consciousness, this thesis shifts the focus to how the appearance of these affects human decision-making. It is argued that LLM AI systems show an appearance of FW (AFW) and an appearance of consciousness (AC) through observable behaviours. This appearance is enough to influence human deliberation, even when users may not believe that AI systems have genuine FW or consciousness. While AI systems do not remove alternative possibilities (APs) or take away human control, they do reshape how options are perceived and evaluated, which narrows the phenomenological experience of freedom. This influence often operates below conscious awareness, affecting reasoning processes without coercion. Traditional views of FW and moral responsibility that focus on the ability to do otherwise do not adequately account for external influence on deliberation from AI systems. The thesis concludes that responsibility and accountability should remain with humans, as AI systems function as a tool rather than morally responsible entity. However, the increasing use of AI in everyday decision-making suggests the need to reconsider how influence, human cognition, responsibility, and accountability interact in an AI-mediated environment.