← Back to Journal

Recurring Narrative Evidence Expansion

Tags: voice-analysis • recurring-narratives • corpus • validation

Recurring Narrative Evidence Expansion

Summary

Voice Analysis expanded the recurring-narrative hypothesis from one 2009 time-machine sequence to five human-reviewed clusters spanning 2009 through 2024.

Method Lesson

The evidence unit for recurrence is a relationship between posts.

A repeated word, topic, real event, organization, or platform pattern is not enough. Later writing must continue, modify, recall, or assume familiarity with a distinctive earlier element.

Model retrieval discovered candidates, deterministic search expanded them, and human review rejected false recurrence.

Result

Retained clusters include the time machine, overheard-at-work frame, pay-it-forward inversion, GrandBrad persona, and annual National Daughters Day failure.

EVID-003 and VAL-003 were revised to Version 2. The hypothesis remains provisionally supported with medium confidence inside Facebook writing.

Integrity Lesson

One model citation used the wrong identifier, one run exhausted its output budget, and one returned an empty length-limited result.

Review also found thirty-three residual Mobile uploads Place: records in the unflagged batches. They were excluded manually, but the eligibility rule should be considered for a future correction.

Next Step

Validate and commit this research increment, then address the residual platform-artifact rule as a separate workflow improvement.