State of Evidence for the Readout Hypothesis-Generation Programme: A Shared Evidence Registry for Papers I–IV (with Claim Registry E01–E25)

Series. Companion evidence registry for the four-paper Bounded Knower / Readout hypothesis-generation programme: Paper I (DOI 10.5281/zenodo.22307148), Paper II (DOI 10.5281/zenodo.22307561), Paper III (DOI 10.5281/zenodo.22307564), Paper IV (DOI 10.5281/zenodo.22307751). Files: the State of Evidence synthesis (PDF) and Claim_Registry_E01_E25.csv (25 evidence claims: id, status, evidence strength, primary paper, decisive test). Status. Living GLOSA K0 evidence synthesis; targeted and system-oriented, not preregistered or systematic. "Observed" means the cited literature reported the neighbouring phenomenon, never that a paper-specific Readout variable was validated. No independent checker has certified the registry. No priority claim. Abstract. work of the programme, but it is not itself an observed fact. The four-paper Readout hypothesis-generation programme spans problem formation, bounded semantic mobility, knowledge organization, candidate-set formation, and human–AI amplification. Most local mechanisms invoked by the programme are not new discoveries. Problem representation affects later solving; semantic-memory organization relates to creative and convergent retrieval; domain knowledge changes early diagnostic hypothesis generation; search behavior depends jointly on representation and process; recent inference history biases later inference; expertise can accelerate familiar reasoning while inducing fixation; representational change can resolve impasse; scientific-hypothesis timing is measurable and speed can dissociate from quality; research institutions and incentives route which problems and relations receive attention; scientific communities face unconceived alternatives; and current AI systems can recursively generate, critique, test, revise, and also homogenize or hallucinate hypotheses. What remains unestablished is the programme's proposed integration and its specific observables: readout-weighted accessibility, content-matched first passage to usable hypothesis classes, a measurable live candidate-set operator, and controlled human–AI epistemic multiplication. This document registers the state of evidence claim by claim, separates evidence from synthesis, and identifies the decisive experiments required to move Papers I–IV beyond K0. Keywords: state of evidence; hypothesis generation; scientific discovery; semantic memory; expertise; fixation; candidate-set formation; human-AI collaboration; AI for science; Readout Universe; Readout Genesis • OPEN: a paper-specific claim or variable not yet directly established. • DECISIVE TEST: a record that would materially confirm, narrow, or falsify the open claim. Evidence strength is additionally marked STRONG, MODERATE, or ADJACENT. "Strong" means direct and/or convergent evidence for the local phenomenon; it does not mean causal proof of the full chain. Non-collapse rule for the evidence registry. neighboring evidence ̸= formal-variable validation ̸= truth of the integrated theory The same paper may therefore be strong evidence for a local phenomenon and only adjacent evidence for a Readout-level claim. 2 Master Evidence Map