Comment by 001sky | Hacker News Reader

Comment by 001sky | Hacker News Reader

Parent Full thread

This raises a good point, but informal inferences about age, ethnicity, and sex can be made from other parts of the application. Statistics on the discrimination of Ivy league schools show that (east) Asian face a stacked deck in college applications at "selective" schools[0]. Its plausible that much of this is inferred from ethnic names. To the extent it impacts the short-lists for interviews (ie, before candidates are seen in person), it's obviously detrimental.

Which brings up another point: it would be interesting to see a YC batch where the colleges/universities' names were redacted from the screening process.[1] Again, I don;t think this will ever happen...but as a thought experiment I would likely be of equal interest in terms of "opening" access. At some stage, business is as much about trading favours as it is about measuring "competence". There are some good game-theoretic reasons for this (ie, establishing trust in sequential repeated games), but there is more to the story than that.[2]

___________

[0] http://www.businessinsider.com/ivy-league-discriminates-agai...

[1] Even if this was replaced by a sort of rating system, eg. that placed X schools into N buckets. This could be done so that the information was recorded but never made visible (say by online application). And the data could still be verified later prior as part of due-dilligence/ affadavit to avoid a problem with gaming the system.

[2] http://www.bloomberg.com/news/2013-12-30/jpmorgan-s-mistake-...

The last time I was hiring, I wished I could easily review resumes without seeing the names, email addresses, physical addresses, school names, or even telephone numbers as I felt that I was bringing in my arbitrary bias.

Telephone numbers? I found that I identified a lot more with area code 206 than 425 or 253, just as I identify a lot more with an @gmail.com address than @hotmail.com or @aol.com.

That's a great idea. Next time we put out a request for resumes we might just make something that allows for this. And put it on github..

If you wind up doing that, let me know. I would be willing to help if at all possible. Contact info is in my profile. Thanks.

Don't you think, though, if you have such arbitrary biases, it would be better to work on them than hide the problem?

It wouldn't be hiding biases, it would be making them inconsequential. That's pretty much the best you can hope for, once you admit the possibility that there's no such state as bias-free.

So what, you're going to get all the way to the job offer without knowing their name?

Let's imagine an totally awesome resume from this candidate:

- Kaytlyn (female, youngish, spelled unconventionally)

- who graduated from WSU (the rival of my alma matter)

- who has a 253 area code (my least favorite suburb)

- who uses Papyrus for headings (my least favorite font)

- with the email address belieber69@aol.com (triple yuck)

I would want to get that person in for an interview and explicitly check the subtle biases of me and other people who are making hiring decisions.

True story: when I was in college I had a classmate in my database class who was so good-looking it kind of hurt to look at her. I never once explicitly thought that she was a dumb blonde, but I was surprised when I found out that she was just brilliant. Similarly, I worked on a group project with a few prototypical "frat boys" with their Abercrombie sweaters and backwards baseball caps, and found myself surprised that they were smart as hell, too.

I just want to give people an honest chance to be brilliant and not have their resumes passed over for bullshit reasons, even subconsciously.

I would like to think that I could do both. I really want to believe that the playing field is level no matter if your first name is Paul or Venkatesh or Bambi or LaTonya, but I don't have absolute trust in that. The "screen" is a tool in helping with that.

But how would you do that? Hiding biasing factors is a simple, actionable approach. Do you have a better idea?

> "Statistics on the discrimination of Ivy league schools show that (east) Asian face a stacked deck in college applications at "selective" schools[0]. Its plausible that much of this is inferred from ethnic names."

I wonder how a surname like "Lang" would fair. It is either Germanic of Asian, though it seems to be primarily Germanic in practice but seems strongly Asian to people who are not familiar with it[0]. If there is discrimination keyed off of "Asian-ish sounding" names then it might be apparent when looking at these sorts of names.

[0] I know a germanic "Lang". Apparently he gets asked how his family got that name a lot.

Your comments on school history etc are completely valid in the context of inference.

I'd like to think that we could genuinely make a fairly well balanced system for meritocratic selection. Yes, it's a lot of work and there is always room for error however I'd like to think that ultimately the STEM industry favours these kinds of methods and they could be improved on so we'd see some kind of futuristic system that we saw in the Starship Troopers narrative (as a crazy example that in the movie at least, no-one complained about). Maybe we just need a ton more data to be able to make better predictions. But I also think that face-to-face interviews are ultimately needed as others have mentioned: cultural fit is important to a degree as well.

On that note it reminds me of the Declara article I read (about the founder Ramona Pierson), where data is working to pair relevant people.

https://www.declara.com

Reply on news.ycombinator.com