Preserving Uncertainty across Successive Summaries of Testimony in Institutional Investigations

Institutional investigations often rely on records that have passed through several authors and summaries. A later document can faithfully transmit a mistaken belief, silently strengthen a qualified account, or multiply the apparent evidential weight of one source. This paper develops a typed framework for preserving uncertainty across those transformations. It separates the event, a witness's interpretation, the witness's recorded account, a summarizer's inference, and an authorized finding. A semantic weakening condition characterizes transformations that introduce no new asserted consequences within a fixed interpretation domain. A distinct factorization condition determines exact recovery of declared source targets. Constructed counterexamples show that conservative weakening can still lose qualifications needed for a later inquiry, and that accurate source-target extraction can transmit a false allegation. A probabilistic source-dependence model compares repeated copies with independent reports and derives bounds when their generating structure remains uncertain. An optimal-risk comparison clarifies why information loss can coexist with improved performance by actual readers. A worked institutional packet retains target time, attribution, qualifications, source access and revision history. Primary provenance, archival, source-monitoring and investigation materials establish antecedents and practical constraints. A proposed evaluation separates record fidelity, attribution accuracy, uncertainty recovery and justified belief. The formal results are conditional; institutional effectiveness and the prevalence of the diagnosed failures remain empirical questions.