3 ms·

Is this clunky? Yes. Is there a better way of doing it? Maybe. Is it an explicitly biased policy? Yes. However, correcting a biased system requires that there

by olliej 5y ago

Is this clunky? Yes. Is there a better way of doing it? Maybe. Is it an explicitly biased policy? Yes. However, correcting a biased system requires that there is a period with enforced change bias to correct the existing bias. This is because there are many ways previous explicitly biased policies entrench that bias even if the current environment is no longer explicitly biased. Even if we were to assume that everyone is suddenly 0% biased on anything, a lack of explicit corrective actions will necessarily be long and dragged out continuing to harm the classes that are no longer explicitly discriminated against. So given we know that hiring biases do still exist, and we know that many groups are underrepresented due to historical biases we end up stuck trying to fix a complex problem with very clunky binary tools. As a very simple example: imagine you have a job where the starting rate for one group is $2 and for the other group it’s $1. Everything else is fair people get raises at the same rate (eg 10% or something). Then one day the law says “no, that’s bullshit”, and now everyone’s starting rate is $2. At this point all the people who were already employed are still earning half as much as their colleagues, despite the policies now being completely “fair”. If you just leave things as they are, eventually things will end up equal across both groups, but you have a potentially multiple generation latency. You could tried to correct the incomes, but that looks like discrimination: “we’re doubling the income of everyone in group B, group A is explicitly excluded”. The thing is in this case, even with this biased 2x income, the disadvantaged group is still worse off, due to the years of unfair compensation. If you try to correct for that you end up with another super biased behaviour. What we’re seeing is an attempt to correct historical biases, and any such correction is necessarily biased, because otherwise even a perfectly unbiased continues to harm historical victims.

WkndTriathlete 5y ago
> However, correcting a biased system requires that there is a period with enforced change bias to correct the existing bias. I guess all my calculus and systems theory textbooks were wrong, then. Seriously, though, there are myriad problems with this premise: a) It implicitly assumes that the system itself is biased instead of its inputs. b) It assumes that the system isn't already being corrected based on the first- or second-derivative on the inputs. c) Feedback loops (enforced change bias) can result in nasty exponential on oscillational responses. And in the context of the code itself: d) The code assumes the problem is systemic across all public departments and thus must apply to all of them; e) The code makes suitability to position (which is the point of hiring a professor) secondary to considerations irrelevant to the position. (Plain English: given two candidates, an old, experienced, non-racist white guy who tries to each and encourage all of his students vs. any candidate of one of the protected classes whose biases run toward their protected class, you must hire the biased candidate. And you can't tell if they are biased because the current environment enables that candidate to continue their bias in the current environment, which is to attempt to lift up those protected classes at any cost - if they exclude other protected classes or "white guys" it won't show up as a negative on the demographic statistics assembled.) I agree there is a discrimination problem with certain people, but I don't think the solution is ever going to come from legislation like this. I think the solution has to be grass-roots and be done by gradually retiring out those with generational unconscious bias or by firing those that are outright discriminatory, and by gradually hiring in a younger generation that is (hopefully) less discriminatory. That's a lot more work than making a half-assed law and winning votes because of that law, though.
joannanewsom 5y ago
How do you determine which ethnic groups have been historically disadvantaged and to what extent need correction? Black: Yes Native American: ? Afghan: ? Puerto Rican: ? Hawaiian: ? Japanese: ? Jewish: ? White: No Or do you simply go until the diversity of the position/company matches the diversity of the general population?
- olliej 5y ago
  Determining groups that have been historically disadvantaged is generally not difficult. The part that is difficult is how you correct for that. There are many things you need to deal with. The most obvious ones being * Fields having relatively small populations run into the standard low population vs. overall population statistics * Fields that have historically restricted access to a group can legitimately point to a lack of "qualified" members in the restricted group Neither has an easy solution. The only thing we know is that a failure to actively work on correcting the effect of historical discrimination extends the impact of that discrimination beyond the point when the discrimination "ended". Like I said in the original post this particular example is clunky, but I also don't have any better ideas :-/
  - joannanewsom 5y ago
    > Determining groups that have been historically disadvantaged is generally not difficult. Want to elaborate on how to determine if an ethnic group is disadvantaged? I really do not see what the process would look like, that isn't just comparing population sizes. Otherwise you get into these impossible questions of is anti-semitism worse than hate crimes against asians. Every ethnicity can point to a history of injustice against them. But if the ideas is to favor certain groups, that means there needs to be a way of choosing which are deserving and which are not.

by olliej 5y ago

WkndTriathlete 5y ago

> However, correcting a biased system requires that there is a period with enforced change bias to correct the existing bias. I guess all my calculus and systems theory textbooks were wrong, then. Seriously, though, there are myriad problems with this premise: a) It implicitly assumes that the system itself is biased instead of its inputs. b) It assumes that the system isn't already being corrected based on the first- or second-derivative on the inputs. c) Feedback loops (enforced change bias) can result in nasty exponential on oscillational responses. And in the context of the code itself: d) The code assumes the problem is systemic across all public departments and thus must apply to all of them; e) The code makes suitability to position (which is the point of hiring a professor) secondary to considerations irrelevant to the position. (Plain English: given two candidates, an old, experienced, non-racist white guy who tries to each and encourage all of his students vs. any candidate of one of the protected classes whose biases run toward their protected class, you must hire the biased candidate. And you can't tell if they are biased because the current environment enables that candidate to continue their bias in the current environment, which is to attempt to lift up those protected classes at any cost - if they exclude other protected classes or "white guys" it won't show up as a negative on the demographic statistics assembled.) I agree there is a discrimination problem with certain people, but I don't think the solution is ever going to come from legislation like this. I think the solution has to be grass-roots and be done by gradually retiring out those with generational unconscious bias or by firing those that are outright discriminatory, and by gradually hiring in a younger generation that is (hopefully) less discriminatory. That's a lot more work than making a half-assed law and winning votes because of that law, though.

joannanewsom 5y ago

How do you determine which ethnic groups have been historically disadvantaged and to what extent need correction? Black: Yes Native American: ? Afghan: ? Puerto Rican: ? Hawaiian: ? Japanese: ? Jewish: ? White: No Or do you simply go until the diversity of the position/company matches the diversity of the general population?

olliej 5y ago
Determining groups that have been historically disadvantaged is generally not difficult. The part that is difficult is how you correct for that. There are many things you need to deal with. The most obvious ones being * Fields having relatively small populations run into the standard low population vs. overall population statistics * Fields that have historically restricted access to a group can legitimately point to a lack of "qualified" members in the restricted group Neither has an easy solution. The only thing we know is that a failure to actively work on correcting the effect of historical discrimination extends the impact of that discrimination beyond the point when the discrimination "ended". Like I said in the original post this particular example is clunky, but I also don't have any better ideas :-/
- joannanewsom 5y ago
  > Determining groups that have been historically disadvantaged is generally not difficult. Want to elaborate on how to determine if an ethnic group is disadvantaged? I really do not see what the process would look like, that isn't just comparing population sizes. Otherwise you get into these impossible questions of is anti-semitism worse than hate crimes against asians. Every ethnicity can point to a history of injustice against them. But if the ideas is to favor certain groups, that means there needs to be a way of choosing which are deserving and which are not.