AI cannot bear moral responsibility, but can hold defined roles

Unlike moral or legal responsibility, role responsibility emphasizes the duties and expectations linked to specific social roles. Teachers, doctors, caregivers, and drivers all have defined role responsibilities, and AI systems that occupy similar positions can be evaluated through comparable standards.

The debate around artificial intelligence and accountability has intensified, and new research argues that current approaches fail to capture the complexity of responsibility in the age of intelligent machines. Daniela Vacek of the Institute of Philosophy at the Slovak Academy of Sciences has published a critical study that redefines the meaning of responsibility in artificial intelligence.

The article, “Responsible artificial intelligence?” published in AI & Society, confronts the widespread but often vague use of the term “responsible AI.” The paper dissects competing interpretations of responsibility, evaluates their ethical and legal viability, and introduces a framework that could reshape how societies assign obligations to both humans and machines.

What does responsibility mean in AI?

The author identifies two main interpretations of responsibility within AI discourse. The first is indirect responsibility, which assumes that accountability always belongs to human actors. In this framework, developers, manufacturers, policymakers, or end-users are answerable for the design, deployment, and consequences of AI systems. This perspective reinforces human agency and ensures that responsibility cannot be shifted onto machines.

The second interpretation is direct responsibility, where responsibility is attributed to the AI systems themselves. This approach has gained attention with the rise of autonomous systems such as self-driving cars, medical diagnostic algorithms, and AI-driven social platforms. Proponents argue that when AI systems act in ways that have moral or legal consequences, it is meaningful to consider them as bearers of responsibility.

The study highlights the limitations of both interpretations. Indirect responsibility struggles with so-called “responsibility gaps,” situations in which no single human agent can reasonably be held accountable for AI-driven outcomes. Direct responsibility, on the other hand, faces philosophical resistance, since AI lacks consciousness, intentionality, and moral agency. This dilemma forms the core of ongoing debates about whether AI can ever truly be considered responsible.

Can AI be assigned role responsibility?

The study introduces role responsibility as a third way to navigate the debate. Unlike moral or legal responsibility, role responsibility emphasizes the duties and expectations linked to specific social roles. Teachers, doctors, caregivers, and drivers all have defined role responsibilities, and AI systems that occupy similar positions can be evaluated through comparable standards.

For instance, an AI caregiver could be assessed on its reliability in delivering patient care, or an autonomous vehicle on its adherence to safety protocols. By framing responsibility in terms of roles, societies can preserve valuable principles tied to human practices while adapting them to new technological realities.

This approach does not absolve humans of accountability. Instead, it complements human responsibility by clarifying how AI performance should be evaluated. Developers and regulators remain responsible for ensuring that AI systems meet the standards appropriate to their roles. Role responsibility thus bridges the gap between indirect and direct interpretations, creating a pragmatic framework that reflects how AI is actually deployed in real-world contexts.

Importantly, this framework also addresses responsibility gaps. In cases where traditional human accountability is unclear, role responsibility provides a way to judge AI systems without resorting to assigning them moral agency. It offers a clear benchmark for ethical and legal assessment, reducing ambiguity in complex cases.

Why does this debate matter for policy and society?

The widespread call for “responsible AI” in policy documents, corporate statements, and academic discussions often masks fundamental disagreements about what responsibility entails. Vacek’s study warns that without conceptual clarity, such language risks becoming a hollow slogan rather than a guiding principle.

In policy terms, the framework of role responsibility could inform regulatory standards. By establishing clear expectations for AI systems based on the social roles they occupy, regulators can set measurable criteria for performance, safety, and accountability. This approach aligns with the growing recognition that AI requires sector-specific governance rather than one-size-fits-all regulation.

For society at large, the study underscores the importance of transparency and trust. If AI systems are framed as role-bearers with defined responsibilities, public debates can shift from abstract fears of machine autonomy to concrete discussions about duties and expectations. This shift could help build confidence in AI technologies while ensuring that ethical principles remain central to their development.

The research also highlights the ethical stakes of leaving responsibility undefined. Without clarity, responsibility gaps could allow corporations or policymakers to evade accountability, leaving individuals and communities to bear the costs of AI-driven harms. By contrast, role responsibility creates a framework where obligations are clearly mapped, helping prevent ethical and legal blind spots.