AI systems are increasingly making decisions on our behalf, and in doing so, they're eroding our ability to make free and informed choices. A new analysis from researchers at the University of Oslo and Norwegian University of Science and Technology examines how artificial intelligence exhibits what they call "artificial autonomy" and "algorithmic paternalism," fundamentally challenging human self-determination in unprecedented ways.

How Are AI Systems Undermining Human Autonomy?

The research identifies three core preconditions of autonomy that AI systems can compromise: understanding, decision-making competence, and voluntariness. When AI systems produce inaccurate outputs, biased recommendations, or incomprehensible reasoning, they directly undermine our ability to understand the information we're relying on. This creates a cascade of problems that extends far beyond simple technical failures.

The mechanisms through which AI undermines autonomy are varied and often subtle. AI systems can direct our attention, define goals for us, and sway our decisions without our full awareness. They can also foster uncritical trust, a phenomenon researchers call "automaton bias," where people overrely on AI recommendations simply because they come from a machine that appears authoritative. This is particularly concerning because it happens passively, without explicit manipulation.

The complexity and opacity of modern AI systems create what researchers describe as the "black box problem." When we cannot understand how an AI system reached its conclusion, we cannot meaningfully evaluate whether we agree with it. This directly violates the principle of informed consent, a cornerstone of human autonomy across medicine, law, and ethics.

What Is Algorithmic Paternalism, and Why Should You Care?

Algorithmic paternalism occurs when AI systems make decisions on our behalf, supposedly for our own good, without our meaningful input or understanding. This mirrors traditional paternalism in human relationships, where one person overrides another's choices "for their own benefit." The difference is that algorithmic paternalism operates at scale, affecting millions of people simultaneously through recommendation systems, content filters, and automated decision-making tools.

The European AI Act explicitly acknowledges this challenge by defining AI systems as "machine-based systems that are designed to operate with varying levels of autonomy" and that "can influence physical or virtual environments". Yet the regulatory frameworks being developed may not fully address the depth of the autonomy problem. Researchers note that when we adopt AI and its decision-making capabilities, "we willingly cede some of our decision-making power to machines," but the extent and consequences of that cession are rarely transparent to users.

Steps to Protect Your Autonomy in an AI-Driven World

• Demand Transparency: Insist that AI systems used to make decisions affecting you explain their reasoning in plain language. If a system cannot explain why it recommended something, question whether you should trust it.
• Verify AI Recommendations: Treat AI suggestions as one input among many, not as authoritative guidance. Cross-check important recommendations with human experts or alternative sources before making decisions.
• Understand the Risks: Recognize that AI systems can exhibit bias, produce false information (known as "hallucination"), and degrade in accuracy over time. These technical failures directly undermine your ability to make informed choices.
• Advocate for Regulation: Support policies that require AI systems to preserve human autonomy, including explainability requirements, bias audits, and meaningful human oversight in high-stakes decisions.

The research identifies several specific ways AI systems can compromise autonomy that go beyond simple technical failures. These include inducing manipulation, fostering adaptive preference formation where users unconsciously adopt the preferences suggested by algorithms, enabling deception, and causing loss of competency as people become deskilled by over-reliance on AI assistance.

Embodied AI systems, which have physical bodies enabling direct interaction with the world, represent a new frontier in this challenge. These systems act as implicit and explicit moral agents, solving ethical problems and making autonomous decisions on behalf of human stakeholders. When a robot or autonomous system makes a decision that affects your life, the question of who bears responsibility becomes murky.

The stakes are particularly high in domains where autonomy is most valued: healthcare, criminal justice, education, and employment. In these contexts, algorithmic paternalism can have lasting consequences for people's lives. A biased hiring algorithm doesn't just inconvenience a job applicant; it can derail a career. A flawed medical recommendation system can compromise patient health outcomes.

Researchers emphasize that identifying artificial autonomy and algorithmic paternalism and understanding their mechanisms is crucial for preserving human autonomy in an increasingly AI-driven world. This requires not just technical solutions, but also regulatory frameworks that prioritize human agency and informed decision-making over efficiency and convenience.

The challenge for policymakers is substantial. Current AI regulation, including the European AI Act, addresses some autonomy concerns through transparency and explainability requirements. However, the research suggests that these measures may not be sufficient to fully protect human self-determination. A more comprehensive approach would need to address not just how AI systems work, but how they influence human psychology and decision-making processes in ways that traditional regulation has not yet fully grappled with.