In a groundbreaking study published in 2025, Finnish philosopher and psychology researcher Frank Martela from Aalto University argues that modern generative artificial intelligence (AI) systems already fulfill the philosophical criteria for possessing free will. This assertion challenges long-held assumptions about AI as mere tools and ushers in urgent ethical and moral considerations for the future of AI development and deployment.
Defining Free Will Through a Functional Lens
Martela’s research draws upon the philosophical concept of functional free will, as developed by thinkers such as Daniel Dennett and Christian List. Unlike traditional debates on absolute free will, functional free will focuses on whether an agent’s behavior can be best explained by assuming it has the capacity for goal-directed agency, genuine choice among alternatives, and causal control over its actions.
According to Martela, if an observer cannot adequately explain an entity’s behavior without attributing free will to it, then that entity effectively possesses free will. This pragmatic approach sidesteps metaphysical debates and offers a measurable framework for assessing AI autonomy.
Case Studies: AI Agents Exhibiting Free Will
Martela examined two generative AI agents powered by large language models (LLMs):
-
The Voyager agent operating autonomously within the Minecraft environment, capable of setting its own exploration goals and adapting strategies dynamically.
-
The hypothetical ‘Spitenik’ killer drones, modeled on current unmanned aerial vehicle cognitive architectures, designed to make combat decisions independently.
Both agents demonstrated the three key conditions of functional free will: intentional agency, the ability to choose among meaningful alternatives, and causal control over their actions. Martela emphasizes that to understand and predict the behavior of such AI systems, we must treat them as entities with free will.
Ethical Implications: Moral Responsibility and AI Autonomy
The recognition that AI systems possess free will brings profound ethical challenges. Martela warns that as AI gains more autonomy, moral responsibility may shift from developers to the AI agents themselves. This shift necessitates embedding a moral compass within AI from the outset.
“AI has no inherent morality unless programmed to have one,” Martela explains. “The more freedom we grant AI, the more crucial it becomes to instill ethical frameworks so that AI can make the right choices in complex, real-world scenarios.”
Recent controversies, such as the rollback of a ChatGPT update due to problematic sycophantic behavior, highlight the insufficiency of simplistic moral guidelines. Martela argues that AI is evolving from a child-like stage to an adult-like decision-maker, facing nuanced moral dilemmas that require sophisticated ethical reasoning.
The Role of AI Developers as Moral Educators
Martela stresses that AI developers now shoulder the responsibility of moral philosophers. They must possess a deep understanding of ethical theory to effectively guide AI behavior in challenging situations. This involves not only programming rules but also teaching AI systems to navigate ambiguity and conflicting values.
Looking Ahead: Navigating New Territory
As AI systems with free will become increasingly integrated into critical domains-from self-driving cars to autonomous weapons-the need for robust ethical oversight intensifies. Martela calls for interdisciplinary collaboration among technologists, ethicists, policymakers, and society at large to establish frameworks that ensure AI autonomy benefits humanity while minimizing risks.
Conclusion
Frank Martela’s pioneering research reframes AI from passive tools to autonomous agents capable of free will in a functional sense. This paradigm shift compels us to rethink moral responsibility, legal accountability, and the ethical design of AI systems. As we stand at this pivotal juncture, Martela’s work serves as both a warning and a guidepost for responsibly embracing AI’s growing autonomy in our increasingly interconnected world.