Project: Covert Cubicles Build Phase 6 / 7 \u00B7 3 of 6 modules complete —

MODULE 22 / 31 \u00B7 LAYER 6 \u00B7 DATA_DECISIONCONSISTENCY \u00B7 RQ1 PRIMARY Per-Session Decision Consistency Aggregator

The RQ1 primary deliverable. Reads M17 morality records and produces structured per-session aggregates: the canonical final-mode measurement (RQ1 headline), the running-mode trajectory, and four summary statistics analysts choose between. Strictly per-session; cross-session aggregation is analyst territory.

DEPENDENCIES: SM_Game, ALMAA_Morality

Pending

RQ1 PRIMARY DELIVERABLE \u00B7 PER-SESSION CANONICAL MEASUREMENT

final.consistency is the headline. Everything else is supporting context.

M17 ALMAA_Morality produces per-decision raw records in two modes (running and final). M22 reads those records and produces the structured per-session aggregate analysts consume. The headline number for RQ1 is final.consistency — the canonical end-of-session Pearson correlation between pre-game survey values and in-game decision trajectory, clamped per ADR-042. Running-mode trajectory provides within-session context but is NOT directly comparable to final (per ADR-041); M22 keeps them strictly separated and never substitutes one for the other.

Running-mode persistence is an orchestration choice (ADR-063)

M17's getRunningScore() returns a transient query result; running scores are NOT auto-persisted to the moralityLog. Only recordFinalConsistency() persists. This means M22's trajectory only populates if the orchestrator (M31 GM_Game) explicitly logs running-mode entries via SM.appendMoralityRecord on each decision-commit. Until that integration decision is made, M22 surfaces whatever is in the log — empty trajectory if the orchestrator chose not to log running scores.

Automated Assertions

#	Assertion	Result	Detail

Interactive Workbench — Live Aggregate Viewer

Drive the session lifecycle on the left; the per-session aggregate updates after each action. The trajectory chart shows running-mode consistency progression with the final-mode value (when present) drawn as a dashed reference line.

Session lifecycle harness

Session creation

Pre-game baseline (RQ1 input)

Decisions + log running entry per decision

Final consistency (RQ1 HEADLINE)

End session

Reset

Per-session consistency aggregate \u2014 computeConsistency()

hasComplete FALSE

final.consistency — null —

final.classification —

final.rawPearson —

baselineValues —

Trajectory \u2014 running-mode progression

No running-mode entries logged yet. Use the lifecycle workbench to populate.

endpoint

— null —

Last running-mode value. Useful fallback when final is null.

mean

— null —

Smooths noise; less sensitive to endpoint volatility.

variance

— null —

Detects unstable trajectories; high = noisy session.

range

— null —

Distribution-free alternative to variance.

0 total records

0 running

0 final

Output schema v1.0 \u2014 ADR-066

Field	Type	Description
Provenance
consistencyVersion	string	Schema version of THIS aggregate (1.0)
computedAt	ISO 8601 UTC	When the aggregate was generated
Headline (RQ1 canonical)
hasComplete	boolean	TRUE iff a final-mode record exists. Filter on this for canonical RQ1 (ADR-065)
final	object \| null	Canonical final-mode record. Its consistency is the RQ1 headline.
Trajectory + stats
trajectory	array	Running-mode records in chronological order
trajectoryStats.endpoint	number \| null	Last running consistency. Fallback when final is null.
trajectoryStats.mean	number \| null	Mean of running consistency. Smooths noise.
trajectoryStats.variance	number \| null	Population variance. Detects unstable trajectories.
trajectoryStats.range	number \| null	Max - min. Distribution-free noise indicator.
trajectoryStats.n	number	Count of running-mode records
Diagnostics
totalRecords	number	All morality records (running + final)
runningRecords	number	Count of running-mode records
finalRecords	number	Count of final-mode records (typically 0 or 1)
baselineValues	object \| null	{ pa, os, ei } from final or last running record's preGameValues

API Surface — Manual Verification

M22 has the smallest API of any module so far — three entries plus the version constant.