Speaker roster -- Vigo County Oversight Board, 2026-01-29

45 diarized labels total; 33 sparse ones are merged into <other> and are not resolved here. The 12 labels carrying substantial speech are resolved above.

Summary

• Confirmed (3): speaker_15/speaker_13 -- Christopher A. Himsel, superintendent (merged, one person split by diarization); speaker_21 -- Chad Blacklock, outside presenter from Stifel (public-finance consultant).
• Probable (1): speaker_39 -- Carrie, a board member, on one indirect naming event ("Carrie, that's like the third or fourth time...") matched to a distinctive pattern of three repeated finance questions. No vote or characterization is attributed to this name anywhere downstream -- the probable grade never carries those claims (SKILL.md §12b hard limit).
• Role-only (8): the presiding-officer cluster (speaker_2, speaker_9, speaker_10 -- chair, never named by anyone); the contaminated speaker_52 label (mixes Chad Blacklock's own speech with a third party thanking him -- held role_only rather than guessed); and the four-label administration cluster (speaker_57, speaker_59, speaker_63, speaker_95 -- CFO/superintendent-level content, but the "Dr. Hemsell" third-person reference rules out the superintendent, and no self-naming event lands inside the cluster itself).

Open questions for the audit pass

1. Consider re-running diarization on speaker_57/speaker_59/speaker_63/ speaker_95 and speaker_52 -- both clusters show clear mid-sentence grammatical splits, suggesting a boundary-detection problem rather than genuinely different speakers switching mid-thought.
2. If the CFO (Donna Wilson, per self-report from the superintendent at [t=85]) can be independently confirmed against the speaker_57 cluster (e.g. from a future meeting where she self-identifies), backfill this roster and the merge accordingly -- do not guess her into this pass.
3. No agenda was available for this meeting; board-member names beyond "Carrie" and the two procedurally-named-but-unresolved movers/seconders ("Mark," "Heather") were never attached to a substantial-speech label and are out of scope for this roster per the task's closed set of 12 labels.