5 ms·

This is an interesting idea. I've wondered myself about doing something similar for interview processes in an attempt to remove bias. Assigning every candidate

by howeyc 5y ago

This is an interesting idea. I've wondered myself about doing something similar for interview processes in an attempt to remove bias. Assigning every candidate a random number, address them by the number on all correspondence. Remove names from resumes, applications, replace experience on resumes with "random corp 1", random corp 2, random school 1 etc. All communication text based.

actually_a_dog 5y ago
You'd lose an awful lot of signal by only using text communication. In particular, I don't see how you could get any sense of how this person might interact with teammates doing it this way.
- austincheney 5y ago
  At the point of initial candidate filtration (competence, qualification) that doesn’t matter. Not at all. Save the signal bits for after qualification against a baseline is affirmed. That is the only way to move past bias and be objective.
  - actually_a_dog 5y ago
    I don't think you're talking about the same thing the person I replied to is.
bunfunton 5y ago
Ah yes hire your cog for the machine with no regard to team fit
- Angostura 5y ago
  The problem is that 'team fit' can often be a synomymn for 'someone like me'. If there are particular attributes that you want to hire for, those should be in the job/person spec and candidates should be assessed against them.
  - KarlKemp 5y ago
    It is quite blatantly not-even-code for “I like him”. That likability doesn’t necessarily depend on the protected classes, but often will.
kube-system 5y ago
I always do first interviews with cameras off, partially for this reason. Not only does it remove opportunity for bias, but it lets both of us focus on what is relevant to the discussion. I wouldn't go as far as what you're suggesting though, because verbal communication skills, etc, are a relevant skill for the job.
Moodles 5y ago
That honestly sounds like an awful strategy to attract talent. Surely you believe there’s a correlation between going to the best schools and toughest companies and being somewhat productive? To ignore all those sígnals to avoid some bias is throwing the baby out with the bath water. Let’s try the experiment. Two startups short on cash. One hires your way and one does not. Which one do you think is more likely to survive?
- hammock 5y ago
  >an awful strategy to attract talent That's not the point though. The point is to remove bias.
  - kerneloftruth 5y ago
    That's single variable thinking, for sure. We've been living through a lot of that these past 2 years. It's easy to solve for one variable; beyond that, one has to balance things so that primary goals (hiring capable and personable talent) are still achieved.
  - pawelmurias 5y ago
    You can just avoid looking at the resume altogether. A fair roll of the dice removes a lot of bias.
  - arnoooooo 5y ago
    Is removing bias more important than finding the right person ?
    - KarlKemp 5y ago
      Bias is, by definition, not finding the right person.
      - Moodles 5y ago
        What? Ok? This was a pointless circular tangent then? "We want the best talent." "No we don't. We want the to remove bias." "Isn't talent more important?" "Removing bias is finding the best talent." That was like reading a conversation with an AI chatbot. We all agree we want the best talent then, yes?
gwd 5y ago
IIRC Amazon tried something like this when feeding a neural net with resumes; but the neural net still managed to pick up the sexist hiring process by learning what kinds of activities were typical "girl" or "boy" activities. Humans with experience could essentially do the same thing (even subconsciously). I'm sure that kind of redacting names &c would help, but it wouldn't be perfect. https://www.reuters.com/article/us-amazon-com-jobs-automation-insight/amazon-scraps-secret-ai-recruiting-tool-that-showed-bias-against-women-idUSKCN1MK08G https://www.reuters.com/article/us-amazon-com-jobs-automatio...
- fivea 5y ago
  > (...) but the neural net still managed to pick up the sexist hiring process by learning what kinds of activities were typical "girl" or "boy" activities. I'm not sure that's a valid take. It's my understanding that Amazon calibrated their machine learning model to answer the question of "help me find more people like the one we already have", and they just so happened to have within their ranks way more people with a specific academic and professional background, which was happened to refer to be men who did activities dominated by men.
KarlKemp 5y ago
Orchestras famously switched to behind-the-curtain auditions. As expected, it got the same sort of objections as in this thread. It’s famous for the ridiculous differences he policy change made. Women's success rates doubled overnight.
- gaganyaan 5y ago
  The result of those changes are actually not so clear: https://statmodeling.stat.columbia.edu/2019/05/11/did-blind-orchestra-auditions-really-benefit-women/ https://statmodeling.stat.columbia.edu/2019/05/11/did-blind-... Blind auditions seem like a good idea regardless, don't get me wrong. But the headlines talking about 50% increases are probably wrong.
908B64B197 5y ago
> Assigning every candidate a random number, address them by the number on all correspondence. Remove names from resumes It was fashionable in France for some time, and it was even attempted. It backfired however. Researchers realized that hires were less diverse when using anonymized resume. > replace experience on resumes with "random corp 1", random corp 2, random school 1 etc. That would remove one of the best predictors of future performance. Especially in CS, not all internships or schools are created equal. Now, with all the "affirmative action" going on, these should be interpreted with a grain of salt. Asians, for example, need much better scores to get into Ivies than some other minorities. And YouTube was caught trashing all white and Asian applicant’s resumes a while ago [0]. [0] https://www.theverge.com/2018/3/2/17070624/google-youtube-wilberg-recruiter-hiring-reverse-discrimination-lawsuit https://www.theverge.com/2018/3/2/17070624/google-youtube-wi...
- jevoten 5y ago
  As of 2017, 69 percent of Google’s workforce was male, compared to 70 percent in 2014, and 91 percent was white or Asian, a percentage that’s barely changed over three years. - https://www.theverge.com/2018/3/2/17070624/google-youtube-wilberg-recruiter-hiring-reverse-discrimination-lawsuit https://www.theverge.com/2018/3/2/17070624/google-youtube-wi... These articles always pull the same tricks. They invent a new racial category, "white or Asian", to avoid stating that only 61% were white - i.e. slightly underrepresented compared to the 63.7% white 2010 US demographics [1]. Although by 2014 the white % had probably dropped enough that they are neither over- nor under-represented at Google (the 2020 census reports non-Latino whites as only 57.8% of the US - a 5.9% drop in only 10 years [2]). But you wouldn't know this from "91% white or Asian". Maybe they can make the problem go away entirely by saying Google is "32% Black or Asian"? [1] https://en.wikipedia.org/w/index.php?title=Demographics_of_the_United_States&oldid=593966589#Race_and_ethnicity https://en.wikipedia.org/w/index.php?title=Demographics_of_t... [2] https://en.wikipedia.org/w/index.php?title=Demographics_of_the_United_States https://en.wikipedia.org/w/index.php?title=Demographics_of_t...
  - 908B64B197 5y ago
    I wonder who and why they are framing it this way. Especially with admission scandals at top schools, there seems to be an unspoken systemic bias against asians perpetuated by some activists.
Reichhardt 5y ago
They tried this in Australia, and it resulted in more males being hired: https://www.abc.net.au/news/2017-06-30/bilnd-recruitment-trial-to-improve-gender-equality-failing-study/8664888 https://www.abc.net.au/news/2017-06-30/bilnd-recruitment-tri... "The trial found assigning a male name to a candidate made them 3.2 per cent less likely to get a job interview. Adding a woman's name to a CV made the candidate 2.9 per cent more likely to get a foot in the door. "We should hit pause and be very cautious about introducing this as a way of improving diversity, as it can have the opposite effect," Professor Hiscox said."

by howeyc 5y ago

actually_a_dog 5y ago

You'd lose an awful lot of signal by only using text communication. In particular, I don't see how you could get any sense of how this person might interact with teammates doing it this way.

austincheney 5y ago
At the point of initial candidate filtration (competence, qualification) that doesn’t matter. Not at all. Save the signal bits for after qualification against a baseline is affirmed. That is the only way to move past bias and be objective.
- actually_a_dog 5y ago
  I don't think you're talking about the same thing the person I replied to is.

bunfunton 5y ago

Ah yes hire your cog for the machine with no regard to team fit

Angostura 5y ago
The problem is that 'team fit' can often be a synomymn for 'someone like me'. If there are particular attributes that you want to hire for, those should be in the job/person spec and candidates should be assessed against them.
- KarlKemp 5y ago
  It is quite blatantly not-even-code for “I like him”. That likability doesn’t necessarily depend on the protected classes, but often will.

kube-system 5y ago

I always do first interviews with cameras off, partially for this reason. Not only does it remove opportunity for bias, but it lets both of us focus on what is relevant to the discussion. I wouldn't go as far as what you're suggesting though, because verbal communication skills, etc, are a relevant skill for the job.

Moodles 5y ago

That honestly sounds like an awful strategy to attract talent. Surely you believe there’s a correlation between going to the best schools and toughest companies and being somewhat productive? To ignore all those sígnals to avoid some bias is throwing the baby out with the bath water. Let’s try the experiment. Two startups short on cash. One hires your way and one does not. Which one do you think is more likely to survive?

hammock 5y ago
>an awful strategy to attract talent That's not the point though. The point is to remove bias.
- kerneloftruth 5y ago
  That's single variable thinking, for sure. We've been living through a lot of that these past 2 years. It's easy to solve for one variable; beyond that, one has to balance things so that primary goals (hiring capable and personable talent) are still achieved.
- pawelmurias 5y ago
  You can just avoid looking at the resume altogether. A fair roll of the dice removes a lot of bias.
- arnoooooo 5y ago
  Is removing bias more important than finding the right person ?
  - KarlKemp 5y ago
    Bias is, by definition, not finding the right person.
    - Moodles 5y ago
      What? Ok? This was a pointless circular tangent then? "We want the best talent." "No we don't. We want the to remove bias." "Isn't talent more important?" "Removing bias is finding the best talent." That was like reading a conversation with an AI chatbot. We all agree we want the best talent then, yes?

gwd 5y ago

IIRC Amazon tried something like this when feeding a neural net with resumes; but the neural net still managed to pick up the sexist hiring process by learning what kinds of activities were typical "girl" or "boy" activities. Humans with experience could essentially do the same thing (even subconsciously). I'm sure that kind of redacting names &c would help, but it wouldn't be perfect. https://www.reuters.com/article/us-amazon-com-jobs-automation-insight/amazon-scraps-secret-ai-recruiting-tool-that-showed-bias-against-women-idUSKCN1MK08G https://www.reuters.com/article/us-amazon-com-jobs-automatio...

fivea 5y ago
> (...) but the neural net still managed to pick up the sexist hiring process by learning what kinds of activities were typical "girl" or "boy" activities. I'm not sure that's a valid take. It's my understanding that Amazon calibrated their machine learning model to answer the question of "help me find more people like the one we already have", and they just so happened to have within their ranks way more people with a specific academic and professional background, which was happened to refer to be men who did activities dominated by men.

KarlKemp 5y ago

Orchestras famously switched to behind-the-curtain auditions. As expected, it got the same sort of objections as in this thread. It’s famous for the ridiculous differences he policy change made. Women's success rates doubled overnight.

gaganyaan 5y ago
The result of those changes are actually not so clear: https://statmodeling.stat.columbia.edu/2019/05/11/did-blind-orchestra-auditions-really-benefit-women/ https://statmodeling.stat.columbia.edu/2019/05/11/did-blind-... Blind auditions seem like a good idea regardless, don't get me wrong. But the headlines talking about 50% increases are probably wrong.

908B64B197 5y ago

> Assigning every candidate a random number, address them by the number on all correspondence. Remove names from resumes It was fashionable in France for some time, and it was even attempted. It backfired however. Researchers realized that hires were less diverse when using anonymized resume. > replace experience on resumes with "random corp 1", random corp 2, random school 1 etc. That would remove one of the best predictors of future performance. Especially in CS, not all internships or schools are created equal. Now, with all the "affirmative action" going on, these should be interpreted with a grain of salt. Asians, for example, need much better scores to get into Ivies than some other minorities. And YouTube was caught trashing all white and Asian applicant’s resumes a while ago [0]. [0] https://www.theverge.com/2018/3/2/17070624/google-youtube-wilberg-recruiter-hiring-reverse-discrimination-lawsuit https://www.theverge.com/2018/3/2/17070624/google-youtube-wi...

jevoten 5y ago
As of 2017, 69 percent of Google’s workforce was male, compared to 70 percent in 2014, and 91 percent was white or Asian, a percentage that’s barely changed over three years. - https://www.theverge.com/2018/3/2/17070624/google-youtube-wilberg-recruiter-hiring-reverse-discrimination-lawsuit https://www.theverge.com/2018/3/2/17070624/google-youtube-wi... These articles always pull the same tricks. They invent a new racial category, "white or Asian", to avoid stating that only 61% were white - i.e. slightly underrepresented compared to the 63.7% white 2010 US demographics [1]. Although by 2014 the white % had probably dropped enough that they are neither over- nor under-represented at Google (the 2020 census reports non-Latino whites as only 57.8% of the US - a 5.9% drop in only 10 years [2]). But you wouldn't know this from "91% white or Asian". Maybe they can make the problem go away entirely by saying Google is "32% Black or Asian"? [1] https://en.wikipedia.org/w/index.php?title=Demographics_of_the_United_States&oldid=593966589#Race_and_ethnicity https://en.wikipedia.org/w/index.php?title=Demographics_of_t... [2] https://en.wikipedia.org/w/index.php?title=Demographics_of_the_United_States https://en.wikipedia.org/w/index.php?title=Demographics_of_t...
- 908B64B197 5y ago
  I wonder who and why they are framing it this way. Especially with admission scandals at top schools, there seems to be an unspoken systemic bias against asians perpetuated by some activists.

Reichhardt 5y ago

They tried this in Australia, and it resulted in more males being hired: https://www.abc.net.au/news/2017-06-30/bilnd-recruitment-trial-to-improve-gender-equality-failing-study/8664888 https://www.abc.net.au/news/2017-06-30/bilnd-recruitment-tri... "The trial found assigning a male name to a candidate made them 3.2 per cent less likely to get a job interview. Adding a woman's name to a CV made the candidate 2.9 per cent more likely to get a foot in the door. "We should hit pause and be very cautious about introducing this as a way of improving diversity, as it can have the opposite effect," Professor Hiscox said."