It’s similar to how whiteboarding is an inaccurate but useful model of coding ability. I’ve had candidates who could answer in general detail how they would implement an algorithm, but struggle to produce working code.

The purpose of these behavioral questions is to get you to produce details about how you actually act at work in challenging circumstances.

If we reject these models I think we’d be stuck with contract-to-hire as our evaluation model which seems more stressful for both sides.