Beyond Capability: The Missing Dimensions of Artificial Agency

Artificial intelligence has advanced along a recognizable developmental trajectory, progressing from symbolic reasoning to perception, integrated perception–reasoning systems, action, and increasingly sophisticated world models. While this trajectory captures major gains in artificial capability, it obscures a set of foundational dimensions required for agency in a stronger sense. This paper argues that contemporary AI systems lack five constitutive dimensions of agency: affective grounding, lived intuition, ethical judgment, relational responsibility, and second-person orientation. Drawing on phenomenological philosophy, enactivist cognitive science, and care ethics, we demonstrate that affect in AI is treated as an external variable—recognized or simulated—rather than as a pre-reflective field of salience that structures perception, motivation, and concern. Consequently, so-called intuition in AI amounts to statistical efficiency rather than lived significance. Ethical behavior remains constrained by rules and policies, not practical judgment capable of resolving value conflict in context. Relational capacities, finally, are interactional rather than ethical, lacking responsibility, mutual recognition, and enduring commitment to others. We introduce formal criteria for distinguishing orthogonal limits—dimensions that cannot be reached through scaling existing capabilities—from merely difficult engineering challenges. The paper engages substantively with neo-functionalist and illusionist challenges to this position, and explores whether hybrid bio-technical systems might satisfy the proposed criteria for genuine agency. These absences are not future developmental stages but structural boundaries that distinguish artificial competence from genuine agency, with significant implications for AI governance, design, and the ethics of human-machine relations.