Dual-Axis Knowledge Assessment: Epistemic Stability and Effectiveness Domain

Abstract

Generative AI has sharply reduced the cost of producing and disseminating knowledge, forcing knowledge assessment to separate two questions: how reliable is the knowledge claim itself, and does it take effect for this user on this task channel? We propose a two-axis framework. The vertical axis, epistemic stability, is the stability of a claim under systematic challenge by an assessment community—a proposition-level property decoupled from usage channels—for which we offer three falsifiable candidate measures (refutation cost, cross-domain consistency, axiom dependence) and one predictor (historical stability), falsifiability being conditional on the reliability baseline of community judgments (6.5(vi)). The horizontal axis, effectiveness domain, captures how well knowledge works on a concrete (user, model, task) channel, monitored by the Knowledge Effectiveness Index (KHI) = discriminability − unreliability; a schematic channel-theoretic derivation shows KHI is not an approximation of mutual information. The axes are conceptually separable and jointly yield a four-quadrant intervention scheme whose empirical content is currently limited to measurable domains (6.2); their empirical orthogonality is left as a testable hypothesis. We position this paper as a conceptual proposal—a nascent framework, falsifiable candidate measures, and testable components—without claiming a new paradigm.