📖 Content moderation glossary
What is a moderation threshold?
In short
A moderation threshold is the cut-off value applied to a model's confidence score that decides what happens to content: automatic action above it, human review or no action below it.
Thresholds turn scores into decisions. Many systems use two: a high one above which content is hidden automatically, and a lower one that defines the band of doubtful content sent to a person.
Example: a children's brand sets a very low threshold for insults, because its audience is young and its tolerance is minimal. A political news site sets a higher one, because heated debate is part of its purpose and it does not want to remove sharp but legitimate opinions.
Raising a threshold reduces false positives but increases false negatives, and the reverse. There is no universal right value: it depends on the audience, the category of content and the cost of each type of error. OpenModeration adapts moderation to each client's rules and use cases and has the doubtful zone handled by its own team, while you see every decision in real time.