Human Agency in the Age of Artificial Intelligence

Current discourse on artificial intelligence (AI) is often shaped by two limitations. First, AI is framed as a technological competitor to human agency. Second, governance is treated primarily as the control of technical risks. This conceptual article proposes a different perspective. Its central question is not whether human or machine agency increases in the abstract, but how AI redistributes, expands, constrains, and transforms the conditions of human agency. Drawing on the capability approach, social-cognitive theories of agency, structuration theory, self-determination theory, research on work autonomy, and adaptive governance, the article develops a five-dimensional model. Capacity comprises knowledge, skills, and means; Room denotes institutionally, materially, and socially available space for action; Will refers to motivational and volitional activation; Direction concerns the epistemic and normative formation of goals; and Identity captures the self- and role-conceptions from which responsibilities and purposes emerge. The model treats agency as a relational and dynamic configuration rather than an additive personal attribute. Building on it, the article proposes an Agency Impact Assessment that combines distributional analysis, uncertainty mapping, reversible experimentation, and continuous learning. Applications to knowledge work, algorithmic management, education, social media, and political economy illustrate the framework. The article argues that AI governance should be complemented by the deliberate societal governance of the development and distribution of human agency. It concludes with limitations, operationalization requirements, and an empirical research agenda.