From Problem to Hypothesis: Dynamic Semantic Mobility, Bounded Knowers, and the Readout-Discriminable Hypothesis Space

Series. Paper I of the three-paper Bounded Knower sequence. Paper II: Knowledge Topology and the First Passage to Usable Hypotheses (DOI 10.5281/zenodo.22307561). Paper III: Before Evidence Can Decide (DOI 10.5281/zenodo.22307564). Paper IV: The Epistemic Chain Reaction (DOI 10.5281/zenodo.22307751). Status. Bounded-knower final author draft under the Readout Universe / Readout Genesis architecture; GLOSA K0 — timestamped and citable, not independently reviewed, no priority claim. Claim tiers explicit: [definition], [Dr], [Open], [finite_diagnostic]. Produced with the glosa methodology (Rigour Without Infrastructure, DOI 10.5281/zenodo.22301060); the founder's raw lines behind it are in the Blackbox Log (concept DOI 10.5281/zenodo.22302518, e.g. entry BBL-2026-09-04-094 on imagination as reversible transport). Abstract. Scientific hypothesis generation is distributed across literatures on problem construction, abduction, analogy, model-based reasoning, semantic memory and control, insight, underdetermination, experimental discrimination, distributed cognition, and AI-assisted discovery. This paper proposes a readout-native architecture for the narrower domain of hypothesis-generating imagination. Its central thesis is that imagination can be modeled as bounded dynamic semantic mobility: not an infinite faculty and not a static set of reachable ideas, but a time-indexed process over semantic readouts whose transitions are provenance-audited, dynamically weighted, history-sensitive, and explicitly restructurable. Binary reachability is separated from graded accessibility; current constraints are typed by provenance so that forced distinctions cannot be silently erased while posited or borrowed constraints may be openly relaxed or retyped. Readout-semantic attraction and semantic momentum are introduced as finite access diagnostics, but they are explicitly separated from epistemic warrant and truth: a familiar or institutionally reinforced path may be highly attractive without being well supported. Candidate hypotheses are quotiented by reader equivalence under declared domain actions and a finite horizon, yielding a readout-discriminable hypothesis space. The framework further derives a bounded-knower corollary: epistemic standing is indexed by question, domain, time, reader, policy, and available records rather than inhering permanently in a person or social label. It also extends the architecture to collective epistemic systems without positing a group mind: shared records, artifacts, communication channels, and institutional policies define a collective semantic readout, attraction landscape, momentum trace, and warrant process. Three behavioral propositions remain central and falsifiable: attraction can invert into fixation, recent semantic paths can carry momentum into later search, and productive escape can occur through audited restructuring when a dominant basin fails to reduce residuals. A separate hypothesis concerning whether knowledge organization shortens expected time to a usable new hypothesis is deliberately reserved for a subsequent paper. The present manuscript remains a GLOSA K0 architecture pending independent review.