4 ms·

There's actually a way to get that. Use work-hire tests. Work hire tests are anonymous and unbiased. Either candidates can do the work, or they can't. You hav

There's actually a way to get that. Use work-hire tests. Work hire tests are anonymous and unbiased. Either candidates can do the work, or they can't. You have to set it up so that the work-hire test is all that matters, though, which ~nobody does.

CiPHPerCoder 9y ago
I'm familiar with tptacek's writing. :) This will get off-topic quickly, if we allow it to.
- sillysaurus3 9y ago
  In situations where a subthread is already off-topic and destined for the bottom, I try to make a substantive contribution. It's possible to pull out some interesting conversation even from the dregs. It's true that it grows the subthread, but the [-] button exists now.
door 9y ago
except not everyone has the time or interest to do a ton of unpaid work for companies they're applying to
- CiPHPerCoder 9y ago
  > a ton of unpaid work for companies they're applying to Okay, that's a logistical objection to work-sample tests, which is easily rebutted with, "Simply don't assign _a ton_ of unpaid work" to your candidates. The process we use for hiring software developers at my current employer: 1. Clone this git repository. 2. Build a trivial feature (e.g. adding a search feature to an existing blog platform). 3. Send a patch or pull request. Everyone who's being considered gets the same task, and we base our decision off of several factors: Did they implement a working solution? How much new code did they need to create (knowing that the framework does 99% of this already, and is documented)? If so, did they write unit-testable code? Did they write unit tests? Did they find any of the intentional vulnerabilities? It should, realistically, take most people 2-3 hours at most to complete this task successfully. If they're familiar with PHP development, probably 15-20 minutes. Furthermore, it can be completed at their leisure. If that seems unreasonable, contrast it with the cost of taking a day off work to get dressed up and perform an in-person interview during business hours with complete strangers who are scrutinizing you for fitness, with a very high chance of not getting accepted. The burden of work-sample tests shouldn't be on candidates; the burden should lie with the company to ensure they're collecting objective facts about candidates rather than making subjective decisions.
jnty 9y ago
They're not necessarily unbiased or even anonymous (in a meaningful way). The problem which is set might somehow favour applicants from certain backgrounds. The language used might be unnecessarily complex or use local idioms. The comments and variable names might betray the candidate as a non-native (but perfectly proficient) English speaker. If the candidate uses company hardware, the keyboard may be set to a different layout to what the candidate is used to. There are lots of ways any test which (ostensibly) aims to test raw ability can be very biased indeed. A lecturer of mine once told a story about a horrible experience he had during an exam trying to whisper an explanation of the rules of Checkers to a student who was from a country where the game was rarely played - the possibility hadn't even crossed his mind. Declaring the interview style you use as "unbiased" from the start seems like a great way to get complacent and have large amounts of bias creep in unnoticed.
- sillysaurus3 9y ago
  What are some ways to counter this effect?
  - tbrownaw 9y ago
    Be intimately familiar with, and fully accepting of and comfortable with, every culture that applicants may potentially come from. Or, explicitly require / assume that all applicants be fully familiar with your culture. Or, define an industry standard artificial culture - whether implicit or explicit - and require everyone on both sides to be familiar with it. This could include things like suits and golf for execs, hoodies and beer for techies, social justice activism for webdevs, etc...
  - jnty 9y ago
    I'm hardly an expert, though I'd probably suggest focusing on avoiding the four pitfalls I mentioned. Like any hard process, it involves working hard to achieve gradual improvement, learning from others where appropriate. However, I can guarantee that resting on your laurels and calling your process "unbiased" from the start won't work.
noonespecial 9y ago
Your test may work perfectly and prove that the only people who can "do the work" within your company's structure are a narrow demographic of $current_majority. If your company structure/culture is broken your "unbiased" test probably will be too.
wmil 9y ago
That can get you in trouble. Under US labour law, any test where minority candidates don't pass at four-fifths the rate of others is presumed to be racially biased unless proven otherwise in court. Judges don't have to study stats and end up writing most of these laws. https://www.prevuehr.com/resources/insights/adverse-impact-analysis-four-fifths-rule/ https://www.prevuehr.com/resources/insights/adverse-impact-a...

CiPHPerCoder 9y ago

I'm familiar with tptacek's writing. :) This will get off-topic quickly, if we allow it to.

sillysaurus3 9y ago
In situations where a subthread is already off-topic and destined for the bottom, I try to make a substantive contribution. It's possible to pull out some interesting conversation even from the dregs. It's true that it grows the subthread, but the [-] button exists now.

door 9y ago

except not everyone has the time or interest to do a ton of unpaid work for companies they're applying to

CiPHPerCoder 9y ago
> a ton of unpaid work for companies they're applying to Okay, that's a logistical objection to work-sample tests, which is easily rebutted with, "Simply don't assign _a ton_ of unpaid work" to your candidates. The process we use for hiring software developers at my current employer: 1. Clone this git repository. 2. Build a trivial feature (e.g. adding a search feature to an existing blog platform). 3. Send a patch or pull request. Everyone who's being considered gets the same task, and we base our decision off of several factors: Did they implement a working solution? How much new code did they need to create (knowing that the framework does 99% of this already, and is documented)? If so, did they write unit-testable code? Did they write unit tests? Did they find any of the intentional vulnerabilities? It should, realistically, take most people 2-3 hours at most to complete this task successfully. If they're familiar with PHP development, probably 15-20 minutes. Furthermore, it can be completed at their leisure. If that seems unreasonable, contrast it with the cost of taking a day off work to get dressed up and perform an in-person interview during business hours with complete strangers who are scrutinizing you for fitness, with a very high chance of not getting accepted. The burden of work-sample tests shouldn't be on candidates; the burden should lie with the company to ensure they're collecting objective facts about candidates rather than making subjective decisions.

jnty 9y ago

They're not necessarily unbiased or even anonymous (in a meaningful way). The problem which is set might somehow favour applicants from certain backgrounds. The language used might be unnecessarily complex or use local idioms. The comments and variable names might betray the candidate as a non-native (but perfectly proficient) English speaker. If the candidate uses company hardware, the keyboard may be set to a different layout to what the candidate is used to. There are lots of ways any test which (ostensibly) aims to test raw ability can be very biased indeed. A lecturer of mine once told a story about a horrible experience he had during an exam trying to whisper an explanation of the rules of Checkers to a student who was from a country where the game was rarely played - the possibility hadn't even crossed his mind. Declaring the interview style you use as "unbiased" from the start seems like a great way to get complacent and have large amounts of bias creep in unnoticed.

sillysaurus3 9y ago
What are some ways to counter this effect?
- tbrownaw 9y ago
  Be intimately familiar with, and fully accepting of and comfortable with, every culture that applicants may potentially come from. Or, explicitly require / assume that all applicants be fully familiar with your culture. Or, define an industry standard artificial culture - whether implicit or explicit - and require everyone on both sides to be familiar with it. This could include things like suits and golf for execs, hoodies and beer for techies, social justice activism for webdevs, etc...
- jnty 9y ago
  I'm hardly an expert, though I'd probably suggest focusing on avoiding the four pitfalls I mentioned. Like any hard process, it involves working hard to achieve gradual improvement, learning from others where appropriate. However, I can guarantee that resting on your laurels and calling your process "unbiased" from the start won't work.

noonespecial 9y ago

Your test may work perfectly and prove that the only people who can "do the work" within your company's structure are a narrow demographic of $current_majority. If your company structure/culture is broken your "unbiased" test probably will be too.

wmil 9y ago

That can get you in trouble. Under US labour law, any test where minority candidates don't pass at four-fifths the rate of others is presumed to be racially biased unless proven otherwise in court. Judges don't have to study stats and end up writing most of these laws. https://www.prevuehr.com/resources/insights/adverse-impact-analysis-four-fifths-rule/ https://www.prevuehr.com/resources/insights/adverse-impact-a...