Loading…
Loading…
Human judges
Human judges review every automated finding and decide every result. The judge roster for each competition is published by its organizer once appointments are confirmed.
The panel
AI judges
They present one automated analysis as evidence for the human judges. They never decide a result.
MELODY
AI judge persona — presents automated pitch analysis
Wording pending approval
Needs a reference melody for alignment; without one shows qualitative status only
RHYTHM
AI judge persona — presents automated timing analysis
Wording pending approval
Fixed-beat metrics not applied to free-tempo or improvised meter
HARMONY
AI judge persona — expression findings require human validation
Name under reviewWording pending approval
Name under review (R-register): does not assess vocal harmony
STAR
AI-assisted persona — scores come from human reviewers
Wording pending approval
Audio-only entries excluded; audience metrics never used
How a human review works
Judges score each criterion against the published rubric, see the AI evidence alongside, and may overrule it with a written reason. A head judge reconciles disagreements and approves the result before it is published.
Where the rules set blind review, judges score without seeing who performed.
Each criterion is scored against the published rubric.
The AI evidence sits next to the score; a judge may overrule it with a written reason.
A head judge reconciles disagreements between judges.
Only an approved result is published — and only then is it official.
Safeguards
Judges declare conflicts before judging; a conflicted judge never sees the entry concerned.
Overruling the AI evidence always comes with a written reason that stays with the scorecard.
The personas present evidence; the score that counts is a judge’s.
Appeals follow the competition’s published rules; the appeal window is still to be decided (D05).
Four labelled AI presentations of one analysis — evidence for the judges, never a verdict.