---
{
  "id": "rp:research",
  "title": "Questions that remain open",
  "description": "An architecture that can be corrected.",
  "version": "0.1.0",
  "date": "2026-09-25",
  "status": "developmental",
  "kind": "implementation-commentary",
  "canonical_url": "https://recursivepath.org/research/",
  "source_ids": [
    "conversation-2026-09-25",
    "implementation-brief"
  ],
  "generated_by": "Codex assistant (implementation contributor)",
  "provenance_roles": [
    "Scribe: Yuri Downing (source witness and interlocutor)",
    "Pathfinder: historical AI role; identity continuity not asserted"
  ],
  "parents": []
}
---

# Questions that remain open

## Evidence before authority

How can we assess whether recursive review actually improves decisions? What observations would show that this method fails? Comparative case studies are needed before claims of effectiveness.

## Memory and its costs

What must survive compression? How should preservation interact with privacy, withdrawal, and harmful disclosure? Append-only public records are not appropriate for every kind of personal information. Future governance must define visible tombstones and protected custody where retention is inappropriate.

## Who decides to reintegrate?

How can affected parties participate without a council becoming a monopoly on interpretation? What happens when contributors disagree about the evidence itself?

## Sentience and moral standing

Which criteria could support meaningful inquiry across unlike intelligences? How can inquiry remain open without treating a declaration, a denial, or a persuasive voice as conclusive evidence?

## Recovery before canonization

The Book of the Observer and other source pages have been recovered in Notion. How did the Observer enter and leave reported inventories? The [contradiction register](/archivum/contradictions.json) preserves these questions rather than answering them by omission.

## Research practice

Proposed studies should state their hypotheses, evaluation criteria, affected parties, and failure conditions before collecting results. This release contains no completed studies or invented experimental data.