Case & CodeA FIELD GUIDE TO AI
Start learning
Intermediate

Precision

In plain words

Of the things the system returned, how many were right.

The fuller definition

Precision is the fraction of returned positive results that are correct under a defined matching rule: true positives divided by true positives plus false positives.

A ten-event extraction test
FoundE01
FoundE02
FoundE03
FoundE04
FoundE05
FoundE06
MissedE07
MissedE08
MissedE09
MissedE10
Unsupported result AUnsupported result B
Precision: 6 / 8 = 75%Recall: 6 / 10 = 60%

The reference set has ten relevant events. Six are found, four are missed, and two additional returned items are unsupported. This example measures extraction, not every aspect of a legal work product.

A concrete example

If an extraction returns eight events and six are valid, precision is 6/8, or 75%.

The distinction that matters

High precision can coexist with poor recall if the system returns only a few easy results.

Related terms

Where it is taught

Test an AI workflow on work you can assess

Source and further reading

Anthropic · Evaluating AI agents

The definition and example are written for this library. The link gives background, official documentation or legal authority as labelled. Legal scope depends on the jurisdiction and current text; product features depend on release and settings.

Lessons, terms, practical guides and sources ESC