Attention as Distinction Admission in Finite Systems: Capacity-Limited Gating, Boundary Relevance, and Tunnel Vision

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 Attention as Distinction Admission in Finite Systems is the FDS-M1 working paper in the Distinction Theory / Active Finite Distinction Systems series. The paper develops the attention layer of the agency-semantics spine. Building on the formal core of Active Finite Distinction Systems and the M0 agency-semantics paper, it treats attention as the capacity-limited admission of candidate distinctions into an update-relevant channel. A distinction becomes attended when it is admitted through a finite gate, maintained or made recoverable, assigned a verification path or status, and made available for downstream update, action, prediction, coordination, or boundary maintenance. The paper separates attention from salience: salient distinctions may be rejected when encoding, verification, maintenance, or opportunity costs exceed available capacity, while low-salience distinctions may be admitted when they have high causal boundary relevance or high exploratory information gain. The paper introduces candidate distinction streams, admission gates, scalar and vector attention budgets, boundary-gradient attention value, verification load, background scanning, information gain, semantic deficit, nonlinear gate steepening, tunnel vision, attention hysteresis, attention recovery lag, false admission, collective attention, and epistemic pollution as verification-bandwidth saturation. The paper does not replace cognitive neuroscience, rational inattention, predictive processing, transformer attention, or organizational attention theory. It provides a conservative finite-system bridge for analyzing attention as distinction admission under capacity, verification, and resource constraints. The accompanying replication package includes deterministic synthetic normal-form code, figures, and CSV outputs. The simulations illustrate salience-value dissociation, capacity-limited admission, background scanning, verification-limited rejection, deficit-induced tunnel vision, hysteretic recovery, attention failure modes, and collective attention under epistemic pollution. These figures are synthetic normal-form illustrations, not empirical evidence. Scope. FDS-M1 supports later work on value and goal formation, meaning as actionable semantic quotient, trust as delegated verification, AI alignment, scientific error correction, epistemic pollution, institutional attention, and civilization memory.