That's a plausible hypothesis, but it'd be interesting if someone were willing to test it experimentally. Some evidence for the hypothesis could be found if a "blind" YC batch did much worse than a typical YC batch, measured say 3 or 5 years in the future. Of course, with relatively small sample sizes nothing is likely to be proved beyond doubt, but it'd be interesting to know, and the amount of money needed to test it wouldn't be huge, since YC isn't making VC-level investments. Of course, it's not free or risk-free either, so I could see if they weren't willing to test it.

I've long wanted to see in general some more experimental testing of selection variations. What if YC (or some other funder's) candidates were just selected completely randomly from the applications? What if they were selected solely according to some dumb criterion, like take everyone with the most degrees, or the longest CV, or the most GitHub LoC? What if they were selected purely based on the applications (without the dumb-criterion requirement) but without interviews? For a few tens of thousands of $$, someone willing to try those kinds of things out could get some pretty interesting information on how reliable different selection methods are.

My own hypothesis is a negative one: that beyond screening out a few obviously-bad candidates and taking a few obviously-good candidates, the bulk of the YC selection process is randomly related to outcomes, and the YC mentoring/contacts/press/etc., rather than predictive value of the selection process, is the main driver of their generally strong outcomes. But I can't prove that. :)

While that might be worth experimenting, there's a high cost to it. Having high selection standards makes the network (YC's or any other) exponentially more valuable to those already in it. If you add a few not-so-good apples by mistake, there's no going back.

Also, you have to consider how much quantitative and qualitative experience YC has accumulated, the partners are pretty good at telling in a couple minutes conversation if you're a strong founder. This advantage would be lost with blind interviews.

> the partners are pretty good at telling in a couple minutes conversation if you're a strong founder

This is the part I doubt, though, if by "strong founder" you mean "statistically more likely to exit successfully than people selected according to much simpler 'dumb' criteria". These kinds of claims to predictive ability based on un-quantified holistic properties like "experience" rarely hold up under scientific scrutiny.

Your idea is reminiscent of the Rosenthal-Jacobson study, so it might have some basis in reality :)

http://en.wikipedia.org/wiki/Pygmalion_effect#Rosenthal.E2.8...

1.)The fact the female founders are asking for this, tells you that they don't feel on par with the way things are being done today.

2.) Paul has admitted to being susceptible to the Mark Zuckeberg effect, at least he was honest about that and should be respected for the fact that he realizes that. Most VC's i believe also fall into this trap but don't admit it.

3.) Now what are we going to do about this? Shrug our shoulders and just say this or that won't work or get to trying solutions and iterating on that?

>1.)The fact the female founders are asking for this, tells you that they don't feel on par with the way things are being done today.

How many are asking for this? Where are your stats?

You appear to be requesting science in order to justify doing science. I am here to tell you that is silly.

And I'm here to tell you it's not silly, it's very appropriate to do cheap but possibly flawed research first, before diving into an expensive "science" experiment that would cost some multiple of 1/2 a year of many people's lives.

"forgo an in-person interview altogether. Which doesn't seem like a good idea"

I would disagree strongly in that YC has a measurable financial risk of excluding potentially profitable founders solely for meaningless cultural woo woo reasons. For example if some Finnish dudes conduct perfectly normal business transactions nude in the sauna, a prudish American who refuses to participate has an obvious measurable economic loss solely because of irrational cultural woo woo. Now extend that far out example into female communication style.

Now what would work, or at least would be interesting, is having a female partner interview female founders separately from the male partners then study the female partner's impression vs male partner impression. I don't suspect there would be a huge difference; but at least this would be a somewhat more effective way to test the proposed effect. For my ridiculous made up example, you'd need a Finnish partner; probably easier to run this test on the somewhat easier to acquire and categorize male vs female test subjects.

> Blind applications would be great if they were possible, but I suspect they would be as helpful as a blind audition for concert conductors -- i.e., not at all.

While I agree that blind applications would be somewhat tough for startup founders, conducting seems like a bad example. You could fairly easily judge the resulting music without being able to see the conductor.

I'm going to assume you don't have a lot of familiarity with conducting. It often takes an entire season to rehearse with a group in order to produce music that could accurately be 'judged'. And a conductor is about far more than the music -- how is their rapport with the orchestra? What are they like to work with? What kind of an artistic vision do they have for the group, and how do they communicate that? Ultimately, what kind of a leader are they going to be, along 20-odd different dimensions?

Even with world-class orchestras, where performances are regularly put on with guest conductors after only a few hours of rehearsal time, no permanent conductor would ever be hired on the basis of merely listening to their music. It's a leadership position. (Unlike orchestra players, where it really is more directly about musical proficiency.)

Maybe analogies that require programmers to have a lot of familiarity with conducting aren't super helpful here :)

Guest conductors are hardly unheard of, but you're right that it's not a perfect analogy. Still - it'd potentially make for a decent first go to narrow things down, I'd imagine.

Time and time again, studies have shown that people attribute more positive attributes like kindness and honesty to people who are more physically attractive, irrelevant of sex. This holds even when people are explicitly warned beforehand and told to keep their bias in check.