I struggle a lot with the potentially nonzero overlap between
a) "just legislate that being racist/sexist to people is illegal, then have judges interpret" (which is what seems to be simply necessary to actually make progress against the endlessly inventive evil of racism/sexism),
and b) "but also don't simply make everything illegal because then abusers can selectively enforce the law to punish groups they dislike".
> a) "just legislate that being racist/sexist to people is illegal, then have judges interpret" (which is what seems to be simply necessary to actually make progress against the endlessly inventive evil of racism/sexism),
The current rules are that [in specific listed contexts] you are not allowed to treat individuals differently on the basis of [explicit list of characteristics].
That's essentially your (a), except it's not up to interpretation. Pretending that it is up to interpretation is how you get things like what the article is about (in that case by measuring group statistics rather than comparing individual experiences).
The usual argument I see is that group differences in outcome exist, therefore there necessarily must be discrimination somewhere causing those differences. That this discrimination can't be found is proof of how inventive the people doing the discriminating are.
Since these practices are inferred from their effects rather than directly observed, there's no list.
Evaluating a policy based on Disparate Impact is important because it gets rid of the truck-sized plausible deniability of "Well, see, our policies aren't technically or intentionally racist, even though the end result is clearly that racial minorities are excluded/avoided/attacked/whatever. But since we don't specifically and consciously mention race, then we're in the clear!"
It's kind of a corollary to "The Purpose Of A System Is What It Does": The purpose of a policy is who it impacts. Disparate Impact recognizes this.
Imagine a city where the schools woefully underserve minority students. Despite equal budgets etc. there is a persistent gap, probably due to income gaps but that’s not the point. Within the education system there is no one institutional factor to blame.
Now imagine a civil service exam in that same city, entirely race neutral but resulting in a disparity because it tests general skills. The test is de jure racist when all it does is expose the poor education system. Because the education is so poor, the only remediation is to hire unqualified people. So you end up with a deficient civil service in the name of fairness.
There are other ways to examine whether a test or policy is racist.
>Because the education is so poor, the only remediation is to hire unqualified people. So you end up with a deficient civil service in the name of fairness.
I think there's more solutions to that, no? The poor school might make more poor students, but statistically it probably also has some good ones too. So maybe given a basket of roughly equal candidates to a position, one can treat those with a diverse background as having a benefit, rather than looking at their school background and assuming they're unqualified or setting them aside.
If your candidates don't meet the bar then fine, but i'm not convinced that's the default case in life.
I'm not sure how the approach you're proposing addresses the example JackFr offers.
One reason is that the exam still results in disparate outcomes -- it only tests basic skills -- even though it is not in its design or operation discriminatory. It is a good example of a system that has disparate impact simply as a reflection of some other factor, outside of its control.
Your proposal is to combine the exam with a hiring evaluation that weights some factors besides scores on the exam, but "...given a basket of roughly equal candidates to a position..." is doing a lot of work in that proposal. Is that a given? Is that even likely? That depends on the underlying education system, again; and as long as those baskets of roughly equal candidates are rare, the system you propose would also have considerable disparate impact.
Do you reject the paper's claim (that disparate impact liability makes almost everything presumptively illegal, including in situations where no one has done anything wrong), or do you think that there's no way to address plausibly-deniable intentional discrimination with less collateral damage?
I would say that some systems are not designed to be discriminatory against a group, but since the designer blind spot for the problem might still play a part. Systems designed to be "neutral" has to be evaluated against a ideal neutral. If we find discrepancies then the system is fixed until it's as close to neutral as possible.
By ideal neutral is that their outcome aren't that disparate considering the supposed goal of the system: access to X resource, ability to accomplish X, etc.
What you’re describing is “pretext” and it’s a different thing than disparate impact.
You can find liability for pretext even when something is facially race neutral. You have to show circumstantial evidence that the race-neutral rule is being used to deliberately achieve a racist effect. Using circumstantial evidence to infer intent is pervasive in the law. Almost all criminal laws, for example, require showing intent, but obviously almost nobody ever admits their intent. It’s almost always proven indirectly.
Disparate impact is entirely different—it eliminates the requirement of “racism” altogether. Under disparate impact, there can be liability even if there is zero intent to have a racially disparate outcome. It’s like how the University of California eliminated the SAT because of disparate impact, but thousands of professors now want it back: https://www.bloomberg.com/news/articles/2026-08-06/why-a-pus...
Disparate impact has had a very negative effect by eliminating aptitude tests from the economy. As a result, employers have turned to college degree requirements to perform that sorting function. But that vastly increased the costs for ordinary people. Now you need a college degree just to be a receptionist.
I think a) seems very quixotic and causes more problems than it solves.
Instead, I would say that discrimination should probably only be illegal in places where there aren’t enough of the thing (jobs, etc). For example, what about x-only clubs, or ladies’ nights or gyms or etc.
Unfortunately the only way to have everyone "equal", seems to be to make everyone equally stupid. This idea has been explored in fiction, e.g. Harrison Bergeron.
I don't see a publish date on this? As I understand it, there have been recent shifts away from what's described and towards the relevant government agencies wanting more direct evidence of individuals being treated differently.
The policy warned schools that racial disparities in punishments could trigger federal civil rights investigations under a "disparate impact" standard.
Ergo, presumably, if a school had a single purple student. And that purple student was punished for something. Then the entire school had to also be punished for the thing. To not fall afoul of the law. (Or maybe I'm just too autistic to have a nuanced take.)
> Ergo, presumably, if a school had a single purple student. And that purple student was punished for something. Then the entire school had to also be punished for the thing. To not fall afoul of the law. (Or maybe I'm just too autistic to have a nuanced take.)
Then why even post? You just made something up and then spewed it out.
Our industry should be going after the ATS system use and harmonizing its implementation
Should be going after the camera on policy
Should be going after leetcoding, take home interviews, and every subjective metric that isnt able to be verified in its uniform apllicability across candidates
and according to this paper, even if it was uniform lol
I struggle a lot with the potentially nonzero overlap between
a) "just legislate that being racist/sexist to people is illegal, then have judges interpret" (which is what seems to be simply necessary to actually make progress against the endlessly inventive evil of racism/sexism),
and b) "but also don't simply make everything illegal because then abusers can selectively enforce the law to punish groups they dislike".
Tough. Gotta try, though.
> a) "just legislate that being racist/sexist to people is illegal, then have judges interpret" (which is what seems to be simply necessary to actually make progress against the endlessly inventive evil of racism/sexism),
The current rules are that [in specific listed contexts] you are not allowed to treat individuals differently on the basis of [explicit list of characteristics].
That's essentially your (a), except it's not up to interpretation. Pretending that it is up to interpretation is how you get things like what the article is about (in that case by measuring group statistics rather than comparing individual experiences).
> which is what seems to be simply necessary to actually make progress against the endlessly inventive evil of racism/sexism
Can you give examples of how you think there are “inventive” ways people are being racist/sexist?
The usual argument I see is that group differences in outcome exist, therefore there necessarily must be discrimination somewhere causing those differences. That this discrimination can't be found is proof of how inventive the people doing the discriminating are.
Since these practices are inferred from their effects rather than directly observed, there's no list.
Evaluating a policy based on Disparate Impact is important because it gets rid of the truck-sized plausible deniability of "Well, see, our policies aren't technically or intentionally racist, even though the end result is clearly that racial minorities are excluded/avoided/attacked/whatever. But since we don't specifically and consciously mention race, then we're in the clear!"
It's kind of a corollary to "The Purpose Of A System Is What It Does": The purpose of a policy is who it impacts. Disparate Impact recognizes this.
Imagine a city where the schools woefully underserve minority students. Despite equal budgets etc. there is a persistent gap, probably due to income gaps but that’s not the point. Within the education system there is no one institutional factor to blame.
Now imagine a civil service exam in that same city, entirely race neutral but resulting in a disparity because it tests general skills. The test is de jure racist when all it does is expose the poor education system. Because the education is so poor, the only remediation is to hire unqualified people. So you end up with a deficient civil service in the name of fairness.
There are other ways to examine whether a test or policy is racist.
>Because the education is so poor, the only remediation is to hire unqualified people. So you end up with a deficient civil service in the name of fairness.
I think there's more solutions to that, no? The poor school might make more poor students, but statistically it probably also has some good ones too. So maybe given a basket of roughly equal candidates to a position, one can treat those with a diverse background as having a benefit, rather than looking at their school background and assuming they're unqualified or setting them aside.
If your candidates don't meet the bar then fine, but i'm not convinced that's the default case in life.
I'm not sure how the approach you're proposing addresses the example JackFr offers.
One reason is that the exam still results in disparate outcomes -- it only tests basic skills -- even though it is not in its design or operation discriminatory. It is a good example of a system that has disparate impact simply as a reflection of some other factor, outside of its control.
Your proposal is to combine the exam with a hiring evaluation that weights some factors besides scores on the exam, but "...given a basket of roughly equal candidates to a position..." is doing a lot of work in that proposal. Is that a given? Is that even likely? That depends on the underlying education system, again; and as long as those baskets of roughly equal candidates are rare, the system you propose would also have considerable disparate impact.
Do you reject the paper's claim (that disparate impact liability makes almost everything presumptively illegal, including in situations where no one has done anything wrong), or do you think that there's no way to address plausibly-deniable intentional discrimination with less collateral damage?
Most of the criticisms of POSIWID in https://www.astralcodexten.com/p/come-on-obviously-the-purpo... seem equally applicable to your proposed corollary.
I would say that some systems are not designed to be discriminatory against a group, but since the designer blind spot for the problem might still play a part. Systems designed to be "neutral" has to be evaluated against a ideal neutral. If we find discrepancies then the system is fixed until it's as close to neutral as possible.
By ideal neutral is that their outcome aren't that disparate considering the supposed goal of the system: access to X resource, ability to accomplish X, etc.
What you’re describing is “pretext” and it’s a different thing than disparate impact.
You can find liability for pretext even when something is facially race neutral. You have to show circumstantial evidence that the race-neutral rule is being used to deliberately achieve a racist effect. Using circumstantial evidence to infer intent is pervasive in the law. Almost all criminal laws, for example, require showing intent, but obviously almost nobody ever admits their intent. It’s almost always proven indirectly.
Disparate impact is entirely different—it eliminates the requirement of “racism” altogether. Under disparate impact, there can be liability even if there is zero intent to have a racially disparate outcome. It’s like how the University of California eliminated the SAT because of disparate impact, but thousands of professors now want it back: https://www.bloomberg.com/news/articles/2026-08-06/why-a-pus...
Disparate impact has had a very negative effect by eliminating aptitude tests from the economy. As a result, employers have turned to college degree requirements to perform that sorting function. But that vastly increased the costs for ordinary people. Now you need a college degree just to be a receptionist.
I think a) seems very quixotic and causes more problems than it solves.
Instead, I would say that discrimination should probably only be illegal in places where there aren’t enough of the thing (jobs, etc). For example, what about x-only clubs, or ladies’ nights or gyms or etc.
> discrimination should probably only be illegal in places where there aren’t enough of the thing
That's every job everywhere as far as I can tell. There's fewer jobs that people want than people who want them.
Unfortunately the only way to have everyone "equal", seems to be to make everyone equally stupid. This idea has been explored in fiction, e.g. Harrison Bergeron.
I don't see a publish date on this? As I understand it, there have been recent shifts away from what's described and towards the relevant government agencies wanting more direct evidence of individuals being treated differently.
I saw a 2025 copyright at the bottom.
Sure, but that looks like a page footer. If I go to the site home it's the same, so it's probably the last time the site was a whole was updated.
Easily undone by the next administration.
At least some were the result of court cases, so maybe not that trivial to reverse.
web.archive.org first found it on 9/11/2025
https://web.archive.org/web/20250000000000*/https://www.nyuj...
Obama forced something similar on schools:
https://www.ed.gov/laws-and-policy/civil-rights-laws/race-co...
The policy warned schools that racial disparities in punishments could trigger federal civil rights investigations under a "disparate impact" standard.
Ergo, presumably, if a school had a single purple student. And that purple student was punished for something. Then the entire school had to also be punished for the thing. To not fall afoul of the law. (Or maybe I'm just too autistic to have a nuanced take.)
(Trump rescinded it.)
>Or maybe I'm just too autistic to have a nuanced take.
Who is doing the presuming? That sounds like a bad-faith interpretation.
It's bad math. You can't expect statistics to be applied at that scale, and I think courts would have agreed if anyone had tried.
> Ergo, presumably, if a school had a single purple student. And that purple student was punished for something. Then the entire school had to also be punished for the thing. To not fall afoul of the law. (Or maybe I'm just too autistic to have a nuanced take.)
Then why even post? You just made something up and then spewed it out.
Our industry should be going after the ATS system use and harmonizing its implementation
Should be going after the camera on policy
Should be going after leetcoding, take home interviews, and every subjective metric that isnt able to be verified in its uniform apllicability across candidates
and according to this paper, even if it was uniform lol