Active Finite Distinction Systems as a Criterion for Artificial Agency

Official website: distinctiontheory.orgPublic portal for the start guide, papers, claim status, failure registry, prior-art boundary, and citation resources. Canonical GitHub repository:https://github.com/yiningwu-research/Distinction-Theory Active Finite Distinction Systems as a Criterion for Artificial Agency is a public technical manuscript introducing an operational criterion for artificial agency based on Active Finite Distinction Systems (FDS). The manuscript distinguishes passive input-output mapping from artificial agency by requiring active boundary maintenance, durable update participation, causal loop closure, capacity-deficit management, and resource-governed persistence. The central formal object is an FDS tuple: S = (X, E, B, M, Y, A, U, π, ℓ, Φ, P, τ), where B is an operational boundary, M is memory or model state, U is an update operator, ℓ is boundary-maintenance loss, Φ is a resource budget, P is a pruning or perturbation family, and τ is an update timescale. The paper introduces a taxonomy separating passive mappers, minimal causal controllers, adaptive scaffolded systems, strong FDS-agents, and invariant-supported agents. It also proposes operational tests using update ablation, action-to-future-state influence, capacity-deficit estimation, pruning and externalization efficiency, and resource-governed persistence. This document is not a claim that current large language models lack intelligence, nor that scaling is irrelevant, nor that any named architecture cannot become agentic. It is a structural criterion for when a system containing predictive components qualifies as an artificial agent rather than merely a passive mapper. This release is intended as a citable public technical manuscript and architecture-level reference for future FDS-agent, NextNN, memory metabolism, active pruning, and boundary-maintenance work.