The Invisible Denominator: A Multidisciplinary Theory of Algorithmic Deterrence Before the Decision

Algorithmic fairness is commonly evaluated among people who have already reached an automated decision. Yet an institution may shape who reaches that decision by announcing AI screening, imposing digital prerequisites, presenting discouraging eligibility signals, or requiring complex machine-mediated interactions. People who do not apply may leave no decision record, no appeal, and no trace in conventional fairness dashboards. This conceptual study develops the Invisible Denominator Framework (IDF), which distinguishes decision-stage disparity from participation-stage exclusion and identifies five mechanisms through which algorithmic environments can alter entry into opportunity systems. Through a structured, non-systematic synthesis of administrative-burden theory, applicant-reaction experiments, sociotechnical fairness, selective-label research, and institutional AI governance, the paper specifies the boundary conditions under which non-participation should be treated as a potential equity concern rather than automatically as discrimination. It develops seven testable propositions, an upstream access assurance protocol, and a comparative empirical agenda for recruitment, social protection, education, and credit. The central contribution is a change in the unit of fairness assessment: from outcomes conditional on application to the opportunities, costs, perceptions, and constraints shaping who can apply. Because the population of deterred non-applicants is partly unobserved, the framework recommends triangulation, sensitivity analysis, privacy-preserving outreach, and explicit reporting of uncertainty rather than an unsupported single fairness score. The analysis offers a practical route to making participation-stage harms visible without conflating all non-participation with wrongful exclusion.