When a long body of questions is compressed into a summary, something is lost, and something is gained that can quietly acquire authority. This study asks a bounded question: by what criteria can a reader judge whether a compressed representation of questions serves a declared purpose, preserves identifiable content and plurality, and avoids claiming a mandate to represent speakers or settle decisions that no one conferred? The claim is deliberately limited. A question cannot be assessed by length alone: usefulness, source fidelity, and authority to represent are separate properties, and comparing them requires a traceable unit rather than a purely verbal impression of brevity.
The contribution is conceptual, not empirical. The study distinguishes purposes of compression—attention, discovery, comparison, coordination—from final collective decision; separates wording, answer alternatives, presupposition, practical purpose, and speaker commitment without promising a universal identity test; and defines distinct losses of propositions, distinctions, speakers, stakes, uncertainty, and revision options rather than an unexplained total score. It then proposes, rather than validates, a protocol joining purpose declaration, scoped records, plural outputs, support checks, revision, and decision authority. All examples in the study are constructed, and all empirical evaluations are proposals; participant studies remain unperformed.
Methodologically, the study combines analytic distinctions with small fully specified illustrations—a three-person judgment profile, paired cases changing one property at a time, and a transparent tradeoff example—and with deliberately bounded use of a small source base. Shannon's stated boundary, that its engineering formulation concerns the selection and transmission of messages, helps distinguish a communication measure from an account of meaning rather than proving semantics unimportant. The information bottleneck abstract frames compression as preserving information about a specified relevance variable, which is why relevance here requires a named target. Work on faithfulness in summarization separates unfaithful content from ordinary generation quality, and it is used at abstract level only. The PROV data model supplies entities, activities, and agents as a vocabulary for production history, with the caveat that a provenance edge is an assertion, not proof of preserved meaning. The judgment-aggregation impossibility result motivates, but does not substitute for, the study's own toy profile, since its precise assumptions are not audited here. A reported deliberation study comparing AI-generated and mediator-written common-ground statements is treated as a positive comparison worth examining, not as evidence of decision legitimacy.
The counterargument is serious. Concise summaries materially improve access, and demanding every qualification may exclude readers through overload. Documentation can become ceremonial, provenance expensive, ontologies burdensome, and a framework that absorbs every adverse result unfalsifiable. For a limited personal summarizer, user preference may be the appropriate goal. The study therefore states its failure conditions—simpler tools working equally well, review costs rising without gains, tracing failing, dissent becoming ceremonial, corrections never affecting action—and leaves a research ledger rather than a conclusion of completion.