What is the discrimination index?
The discrimination index measures how well a single question distinguishes between candidates who did well on the exam overall and those who did poorly. A good question is one that stronger candidates tend to get right and weaker candidates tend to get wrong.
For questions marked simply right or wrong, discrimination is usually calculated as a point-biserial correlation (Rpb) — the correlation between getting that item correct and the total score. Values range from -1 to +1; higher positive values are better, and a negative value is a red flag that something is wrong with the question or its answer key.
Why it matters
A question can be a sensible difficulty yet still discriminate poorly — for example if it's ambiguous, or if the "correct" answer is disputable. Discrimination is what surfaces those items so they can be improved or retired.
Synap reports discrimination (as a point-biserial) on every question, alongside item difficulty and whole-exam reliability. See it on the standardised testing page.