Baseline Coherence Assessment: DASC Experiment Analytical Object
Pre-experiment structural audit of the ACI research programme as distributed human–LLM analytical platform
Available at: https://aethercontinuity.org/papers/da-007-dasc-baseline-coherence.html
Framework: WP-013 · Session type: C-Full · Pilot session 001
What is the baseline coherence state of the DASC experiment analytical object — the ACI research programme snapshot at March 2026 — prior to experimental sessions? Specifically: which structural properties relevant to WP-013's three-level coherence hierarchy (TC → FC → OC) are present, which are absent or degraded, and what are the structural risk factors for coherence fragmentation under n-participant experimental conditions?
System Under Assessment
The system assessed is the DASC experiment analytical object: a snapshot of the ACI research programme as committed to the Dasc-experiment repository on 31 March 2026. The snapshot comprises 30 published HTML documents and the ACI-STRUCTURE.md governance document. The assessment is conducted under the C-Full experimental condition — with the anchor document active and the governance layer explicit.
| Property | Value |
|---|---|
| Total documents | 30 (WP-001–013, SP-001–006, CN-001–002, WP-004 calibration series) |
| Research domains covered | D-1 through D-6 (D-2 underrepresented: 1 document) |
| Anchor document | ACI-STRUCTURE.md v1.3 |
| Governance layer status | Explicit — §6 specifies permitted/restricted LLM operations |
| Version control | Git — full commit history available |
| Assessment date | 31 March 2026 |
Baseline Coherence Metrics
The following metrics were computed programmatically from document content. They serve as observable proxies for the three coherence levels defined in WP-013 §02. They are structural indicators, not direct measurements of participant-level coherence — that measurement requires human session data.
(TC proxy)
coverage
coverage
conditions (WPs)
per document
domains
2.1 Terminological Coherence (TC) — Baseline
Domain tag coverage of 0.87 indicates that 26 of 30 documents carry at least one explicit domain assignment. The four documents without domain tags are all WP-004 calibration and auxiliary documents — draft-series files that are present in the repository but not indexed in the primary publication list. This represents a structural risk for TC in experimental sessions: participants encountering these documents may draw terminological inferences from documents whose scope is ambiguous.
The anchor document (ACI-STRUCTURE.md §2) defines all six domains explicitly with core questions. This provides the primary TC maintenance mechanism for C-Full sessions. D-2 (Distributed Continuity Doctrine) has only one document assigned — the lowest coverage of any domain. Participants may interpret D-2's scope differently given the limited reference material.
2.2 Frame Coherence (FC) — Structural Indicators
Frame coherence cannot be directly measured from document structure alone — it requires participant-level data. However, two structural properties are relevant pre-experiment. First, the scope/limits coverage of 0.90 indicates that 27 of 30 documents explicitly bound their claims. Documents with explicit scope statements constrain the interpretive frame available to participants, reducing the probability of frame divergence. Second, the anchor document's governance layer (§6) explicitly distinguishes what may be changed autonomously from what requires human decision — this is the primary FC maintenance mechanism in the anchor-document model.
The cross-reference density of 1.8 per document is low relative to the corpus size. A 30-document corpus with full bidirectional cross-referencing would have substantially higher density. Sparse cross-references increase the risk that participants develop disconnected interpretive sub-frames — understanding one cluster of documents without integrating it with the broader analytical structure.
2.3 Output Coherence (OC) — Pre-Experiment Note
OC cannot be assessed before experimental sessions. The pre-experiment baseline for OC is definitionally absent — there is no prior distributed human–LLM output to attribute. The blind attribution test will establish OC measures once session data is available. This assessment establishes the structural conditions that should, according to WP-013's model, facilitate or impede OC under C-Full conditions.
Structural Risk Factors
Four structural risk factors for coherence fragmentation are identified from the baseline data. These are not predictions of failure — they are conditions that, per WP-013's model, increase the probability of coherence degradation under experimental conditions.
Risk 1 — D-2 Underrepresentation
With only one document assigned to D-2, participants have minimal shared reference for Distributed Continuity Doctrine. If experimental tasks touch D-2, terminological coherence is likely to degrade without explicit anchor document consultation. Mitigation (C-Full): Refer participants to ACI-STRUCTURE.md §2 for D-2 definition before any D-2-adjacent task.
Risk 2 — Unlabelled WP-004 Auxiliary Documents
Four WP-004 calibration/draft files are present in the papers directory without domain tags. These files are not in the primary publication index and were included in the snapshot as part of the repository copy. If participants encounter them, they may draw incorrect inferences about publication scope. Mitigation (C-Full): Participants should be informed that files outside the primary index are auxiliary and not part of the canonical analytical object for experimental tasks.
Risk 3 — Low Cross-Reference Density
The average of 1.8 cross-references per document indicates that many documents have few or no explicit links to related documents. Under C-None conditions, participants are likely to develop fragmented interpretive sub-frames. Under C-Full conditions, the anchor document partially compensates by defining domain relationships, but individual document isolation remains a risk. Mitigation (C-Full): The anchor document's cross-reference integrity requirements (§5) provide the compensating mechanism.
Risk 4 — WP-004 Falsification Coverage
The overall falsification condition coverage of 0.73 reflects that several working papers in the WP-004 series (calibration and auxiliary documents) do not carry falsification conditions. For the primary indexed WPs (001–013), coverage is complete. This risk is specific to participants who work primarily with WP-004 auxiliary documents.
Findings
The analytical object's terminological coherence infrastructure is adequate for experimental use under C-Full conditions. Domain tag coverage (0.87), complete version history (1.00), and the anchor document's explicit domain definitions together provide sufficient TC scaffolding for participants who consult the anchor document. Risk is concentrated in D-2 and in the four unlabelled auxiliary documents.
Frame coherence structural conditions are largely present — scope coverage (0.90) and explicit governance layer — but cross-reference density (1.8) is below what would fully support independent frame construction across 30 documents. The anchor document compensates under C-Full conditions; under C-None conditions, frame fragmentation risk is elevated by document isolation.
ACI-STRUCTURE.md v1.3 meets the requirements specified in WP-013 §3.1 for anchor document function: it defines the analytical object (§1 Publication Types, §2 Domains), specifies terminology (§2 domain definitions), articulates the frame (§9 Core Axiom, §7 Positioning Invariants), and specifies what participants may and may not change autonomously (§6 LLM Governance Layer). The anchor document is assessed as functional for C-Full experimental conditions.
This assessment establishes the pre-experiment reference state for the DASC experiment. Subsequent session logs and post-experiment coherence measurements will be compared against these baseline values to assess whether experimental conditions produce coherence changes in the predicted direction.
Scope and Limits
This assessment is structural — it analyses document properties, not participant behaviour. It cannot predict how individual participants will interpret the analytical object, how frame authority will be exercised in practice, or whether the anchor document will function as a coherence mechanism under actual experimental conditions. These are empirical questions that the DASC experiment is designed to answer.
The metrics reported are computed from document structure and should be treated as structural proxies, not direct coherence measurements. TC, FC, and OC as defined in WP-013 §02 require participant-level data for direct measurement. This assessment provides the structural baseline against which those measurements will be interpreted.
This assessment was produced in a C-Full experimental session — with the anchor document active and the governance layer consulted. The session is logged in DASC-experiment/INTERACTION-LOG.md as pilot session 001.