> but rather that it worked out in practice to be a fairly good filter for whether people actually had (enough of) that skillset. How did you rate your negativ

> but rather that it worked out in practice to be a fairly good filter for whether people actually had (enough of) that skillset. How did you rate your negatives to judge the filter was correct? You know the "Positive" status, you know your true and false positives from the interview question (People you hired that performed well or didn't. Or said another way, the predicted value was positive while true value was positive or negative). But did you know your true and false negatives (the people you didn't hire that would have been good or bad. Or said another way, the predicted value was negative but true value was positive or negative)? Did you know the candidates you rejected who would have been good at fulfilling the roll at your company (False Negatives)? To be clear, I don't know the false negatives either and the interview question may be the best litmus test for the behavior desired, minimizing false negatives and maximizing true positives. But I just don't know for sure if the question did work out in practice to be a good filter of people with that skillset without a true negative dataset. The Precision is good, but the recall/sensitivity may not be. Then again, you probably know how the candidates you hired performed before you used the new question and after - which may be an indicator on false negatives. Also, that does seem like a really fun interview question to answer for me. Kudos. EDITED: Changed the question at the same time the response came from OP. Original post asked if they hired those who failed the test to know more about the negatives. They did.