The Lab · Learn

How AI actually works

Small pages that run on this device. Read the plain version, move the sliders, watch what changes, then say what you saw. Tap any word you do not know.

Lesson 7

Precision and recall

Any system that flags things has to choose between catching everything and only flagging what is truly a match.

It is like: Like a smoke alarm. Too sensitive and it goes off for toast; not sensitive enough and it misses a real fire.
Why it matters at work: Spam filters, fraud checks and screening tools all make this trade. Ask which mistake costs you more before setting one up.

The exact idea

A threshold trades catching everything (recall) against being right about what you flag (precision).

Words on this page

Try this

Slide the threshold to 0.9. Precision is high, but most spam gets through.
Slide it to 0.1. Every spam message is caught, and so are plenty of good ones.
Find where the balance score is highest. Is that where you would set a real filter?

Change it

Ask for a line under the matrix that says how many good messages would land in spam per 100 received.

What happens to precision and recall as you raise the threshold, and why?

Pick the one that fits what you saw.

Both go up, because the filter gets stricter.Recall goes up and precision goes down.Precision usually rises because fewer good messages are flagged, and recall falls because more spam slips under the line.Neither moves. Only the balance score changes.

Field notes

Reading the notes.