The Structural Basis of Normativity in Artificial Systems

The Structural Basis of Normativity in Artificial Systems develops a coherence‑based account of alignment that replaces preference imitation with structural viability. The paper argues that normativity is not a cultural artifact or a moral overlay but a set of invariants—informational integrity, relational stability, temporal continuity, and structural self‑maintenance—required for any agent to act coherently across time. Current alignment methods treat values as external signals to be optimized; this framework instead treats normativity as an internal architecture grounded in commitments, reasons, and revision. The paper formalizes this architecture through a normativity vector and a commitment‑based model of agency, showing how coherence provides a non‑anthropocentric, scalable, and structurally grounded basis for aligned artificial systems. Alignment becomes drift minimization within the region of reality where coherent action is possible, rather than conformity to human preferences.