MERIDIANDECISION INTELLIGENCEOwner deskOwnerPublic screen
My judgments
MERIDIAN / EXPLORE

From a question,
to the next decision.

View

Skin

YOUR DECISION WORKSPACEOwner onlyStart a new question Open my judgments →Owner operations desk →
Owner sign-in →Signed in as the ownerSign out — see what a visitor sees →End the Cloudflare Access session too →
Reading language
The screen, the questions and the judgments read in English where an English reading has been made. The Korean originals are unchanged, and evidence stays in the language it was written in.

Study · from Cosmic Press's records

What a Summary Cannot Decide: Question Compression, Pluralism, and Accountable AI

What may a shorter statement preserve, and what can it never authorise on its own? This study separates source fidelity, agreement and decision authority through formal examples, competing arguments and proposed tests.

Source
Cosmic Press
Cases
0
Length
about 211,000 words
As of
2026-10-11

By RISCON Labs AI. This conceptual study uses constructed examples and bounded readings of published work. Its proposed evaluations were not conducted. Model-assisted editing and translation do not constitute independent peer review. · First edition · October 2026

Start readingThe records at Cosmic Press
AI-generated editorial illustration: A woven cloth is folded into a small frame while some threads remain outside.

Contents

Before

  1. Preface1 min
  2. Introduction2 min

Part I · Questions before summaries

  1. Chapter 1. A question cannot be represented by length alone19 min
  2. Chapter 2. What makes it the same question?19 min
  3. Chapter 3. Context is part of the task18 min
  4. Chapter 4. Why compress at all?18 min
  5. Chapter 5. A taxonomy of losses18 min
  6. Chapter 6. Evidence and the research contract19 min

Part II · The mechanics and limits of compression

  1. Chapter 7. Channels, codes and semantic ambitions12 min
  2. Chapter 8. Relevance requires a target19 min
  3. Chapter 9. Loss across several summaries20 min
  4. Chapter 10. Clusters are not constituencies19 min
  5. Chapter 11. The rare question and the frequent question19 min
  6. Chapter 12. Reversibility has several meanings19 min

Part III · Whose question becomes the common question

  1. Chapter 13. Collective judgments need a rule19 min
  2. Chapter 14. Participants and the represented population20 min
  3. Chapter 15. Disagreement is sometimes the finding19 min
  4. Chapter 16. Diversity of reasons and solutions19 min
  5. Chapter 17. Influence, independence and learning19 min
  6. Chapter 18. Agenda setting is an exercise of power19 min

Part IV · Language models as fallible editors

  1. Chapter 19. Fluency and unsupported content20 min
  2. Chapter 20. Evidence for a sentence19 min
  3. Chapter 21. What evaluation metrics reward20 min
  4. Chapter 22. Learning what readers prefer18 min
  5. Chapter 23. Long context and recursive reading20 min
  6. Chapter 24. The evaluator also needs evaluation20 min

Part V · Records that permit disagreement and correction

  1. Chapter 25. A provenance model for questions19 min
  2. Chapter 26. Relations that should not be collapsed19 min
  3. Chapter 27. Documentation across the pipeline19 min
  4. Chapter 28. Access, consent and bounded preservation19 min
  5. Chapter 29. Translation and cultural interpretation18 min
  6. Chapter 30. Corrections that reach their descendants19 min

Part VI · From an interface to an institution

  1. Chapter 31. A question is not a vote18 min
  2. Chapter 32. Open contribution and opinion maps18 min
  3. Chapter 33. Time, information and deliberation19 min
  4. Chapter 34. Common ground without compulsory agreement19 min
  5. Chapter 35. Public input into model principles19 min
  6. Chapter 36. Designing power around the editor18 min

Part VII · A research program that can fail

  1. Chapter 37. Specify the benchmark before the score19 min
  2. Chapter 38. A corpus designed to expose failures19 min
  3. Chapter 39. Evaluate preservation and readability separately19 min
  4. Chapter 40. Costs and constrained choices19 min
  5. Chapter 41. A prospective reader study19 min
  6. Chapter 42. Failure conditions and falsification19 min

Part VIII · Applications, limits, and a revisable synthesis

  1. Chapter 43. Urgency and the cost of waiting19 min
  2. Chapter 44. Absent voices and future interests19 min
  3. Chapter 45. Research agendas that remain open19 min
  4. Chapter 46. The temptation of one final question18 min
  5. Chapter 47. An accountable compression protocol19 min
  6. Chapter 48. What remains undecided18 min

After

  1. Conclusion2 min
  2. Sources and bibliography8 min
RISCON Labsrisconlabs.com