Why near-ties shouldn't be auto-decided
When a match score lands too close to the confidence threshold to be confident either way, forcing an automatic accept or reject turns genuine uncertainty into a coin flip. A review queue exists to replace that coin flip with human judgment.
What the reviewer actually needs to see
A useful review interface surfaces the original confidence score, the specific comparison that triggered the flag, and enough context to make a real decision — not just a binary flag with no supporting detail.
Closing the loop
The reviewer's decision should feed back into the same audit trail as every automated outcome, so a reviewed case is traceable the same way an automatically-resolved one is — not a separate, less-documented path.