Decoder. plain-English AI glossary

Confidence interval

● Core

A range around your metric that's likely to contain the true value, given your sample.

Think of it like

A fishing net that you're 95% sure contains the fish.

Example

Your model scores 85% accuracy on 1000 examples. The 95% CI is [82%, 88%]. You're saying: if I ran this experiment again, the true accuracy is probably between 82% and 88%.

How it actually works

Confidence intervals are more informative than point estimates. A metric of 85% ± 3% tells you the result is stable. A metric of 85% ± 10% is wobbly and might change on a different sample. The width depends on sample size and variance. Smaller sample = wider interval.

For product teams

Report CIs alongside metrics. They show how confident you should be.

For engineers

Compute from bootstrap resampling or analytical formulas. Report 95% or 90% CIs.

Statistical reporting.

Read anything AI without the jargon

Look up any term in plain English, or save terms as you read with the free Chrome extension.

Open Decoder Add to Chrome