[pmiller2]

I agree. Fairness in the sense of “equal ability to do the job results in the same outcome” would, indeed, get us there. The problem is that interviews don’t actually represent ability to do the job, not in this industry, at least.

[wellpast]

They must to a significant degree or businesses would tank.

The (open) question is, can we do better?

[pmiller2]

I’m on a team with 8 other people right now. All 9 of us are quite capable at our jobs. We all got through some sort of interview process to get there, of course, but that just means we weren’t weeded out. We weren’t false positives, in other words.

But, why did we get through? Maybe the process really is capable of separating wheat from the chaff. Maybe the process works 85% of the time, and the 1 time it didn’t work in recent memory is someone who has already left the company. Maybe only capable people self select into the process for some reason. Maybe we’ve been phenomenally lucky.

The point is, just because a filter allows the right people through doesn’t mean it’s necessarily any good.

[wellpast]

> The point is, just because a filter allows the right people through doesn’t mean it’s necessarily any good.

I define “good” and “right people” to be that which produces efficiency, quality, innovation in the business. So a good filter to me is good wrt these outcomes.

The industry is starving for ‘good people’ and ‘good outcomes’ and we should work to optimize this imo.

Any other measure divorced from outcome to me seems... weird.

[pmiller2]

I wish I could talk about our recruiting funnel numbers. It is the farthest thing from efficient.

Triplebyte claims the industry average onsite pass rate is in the neighborhood of 30% (https://triplebyte.com/blog/12-000-engineers-evaluated). I suspect that the huge majority of those who fail onsites get jobs. Is that efficient?

[wellpast]

I’m advocating for improving that efficiency if it is indeed improvable (my intuition says it is by a large margin.)

What I’m saying is that the notion of ‘fairness’ should have little relevance to improving this side of things.