Hacker News Reader

Why do people always go on to the abstract cases when we are discussing a specific example?

> Where exactly does it stop?

It stops in cases when it's not protected and not breaking the law. Housing and job adverts cannot be explicitly descriminatory. It's a clear existing law. There is a clear line. You have to do your best and take precautions to avoid descrimination. When a platform doesn't do that then there's a problem.

> It stops in cases when it's not protected and not breaking the law.

Since we're discussing the scenarios that the law might cover, your claim seems rather circular.

> It's a clear existing law. There is a clear line.

The entire point of my post is that the line isn't neat and clear at all. Very similar concerns come up in various other contexts, and we as a society are going to have to deal with the internal contradictions of current orthodox thinking one way or another, and probably soon.

Your comment is just asserting that the problem I described doesn't exist. It does. Wishing away reality doesn't make it disappear.

Sure, but the point is Facebook should be taking minimum steps to actively avoid descrimination. It currently is not. Perfect shouldn't be the enemy of the good and they can do better.

I'm not sure people understand the inevitable progression here.

Step 1: HUD files a complaint against Facebook for allowing ad targeting based on certain criteria.

Step 2: Facebook bans advertising targeting based on the criteria in the HUD complaint.

Step 3: Advertisers react by targeting based on demographic correlates of the categories from step #1.

Step 4: HUD notices that ads are still getting delivered more to one group than another.

GOTO STEP #1.

There's no clear point at which this process stops. Demographic correlates are numerous and strong. Are you going to ban all of them? That's tantamount to banning ad targeting generally!

So what about your argument means they shouldn't handle the explicit descrimination cases on their platform better? Just because there might be a fuzzy case in they future why shouldn't they handle the clear cases now?

FYI these are the specific points they are going after:

-display housing ads either only to men or women;

-not show ads to Facebook users interested in an "assistance dog," "mobility scooter," "accessibility" or "deaf culture";

-not show ads to users whom Facebook categorizes as interested in "child care" or "parenting," or show ads only to users with children above a specified age;

-to display/not display ads to users whom Facebook categorizes as interested in a particular place of worship, religion or tenet, such as the "Christian Church," "Sikhism," "Hinduism," or the "Bible."

-not show ads to users whom Facebook categorizes as interested in "Latin America," "Canada," "Southeast Asia," "China," "Honduras," or "Somalia."

-draw a red line around zip codes and then not display ads to Facebook users who live in specific zip codes.

Pretty damn clear: gender, race, disability and religion.

A lot of really bad policy gets enacted because advocates don't honestly ask themselves, "What are the foreseeable consequences of the policy I'm advocating?" That is, a failure to ask the question "And then what?" results in a lot of unnecessary misery.

Your comment essentially restates the HUD's complaint. It doesn't engage with my discussion of the fundamental incompatibility of the HUD's regulatory regime and ad targeting in general. Nobody was ever convinced of anything by someone typing "FYI: [restatement of original premise]".

My point was your premise is making up imaginary scenarios that aren't actually happening as a way of creating an issue with the clear problems outlined and this isn't helpful.

Creating an imaginary general problem as a way of attacking the specific issue is disingenuous.

More generally, housing discrimination will happen at the individual level in all sorts of subtle and not so subtle ways that can’t be generally stopped. That doesn’t mean you throw your hands up and say that we might as allow any sort of discrimination.

If the intent of policy X is to achieve goal Y and X does not in fact achieve Y, what is the point of continuing to pay the costs of X instead of proposing some X' that actually does achieve Y? Continuing to advocate X when X doesn't achieve Y suggests that the proponents of X don't really care about Y, but instead some unstated Z.

Because a simple and easily enforced X that hits 90% of cases is vastly preferable to the 1200 page tome that makes up X’ and requires 175k people to enforce after 20 years of law writing.

In some situations, "perfect is the enemy of the good" is a reasonable approach. Unfortunately, we're not dealing with a good-but-imperfect filter, but a practically useless filter, one that imposes significant costs on top of being ineffective.

"Weapons of Math Destruction" presents good evidence that 90% is much lower than the actual effectiveness of X.

The other person's reply to you was really this simple:

You are asserting a slippery-slope argument, but this is a situation where there are clear places this doesn't lead. It may be fuzzy instead of a clear line, but there's not actually a slippery slope necessarily. And you need to bear burden to show that there is a slippery slope.

https://en.wikipedia.org/wiki/Slippery_slope

Exactly. Thank you.

The slippy slope here is already in operation. It's the entire rationale behind prohibiting targeting based on zip code: demographics change from zip code to zip code, so the HUD concludes that targeting by zip code is de facto targeting by demographic.

What logical basis is there for stopping at zip code targeting and not applying the same logic to other demographic correlates?

> The slippy slope here is already in operation.

Except there are 40 years of court decisions clarifying how the law should be interpreted.

Most of the case law dates from before internet advertising microtargeting. We're in novel legal territory here.

That said, it's possible that a forward-thinking court has already resolved the question under discussion. Can you point me to case law that discusses the boundary between discriminatory and non-discriminatory ad targeting based on demographic correlates?

The fact it’s an online entity may or may not be novel (I imagine Craig’s List has been sued over this).

I meant my comment to refer to the idea of a slippery slope on what does/does not count as a discriminatory. There is lots of case law on that.

I think it’s safe to say if a court has previous rules you can’t put ‘No X’ or ‘Y only’ in a newspaper ads you can use t as a filter for ad targeting.

It doesn't expose me to relevant ads. I live in Montreal. My last three twitter ads were for a drug (reported as illegal) something about a new york lawsuit against an oil company? And for a pizza place which doesn't exist in my country.

User targeted ads do not produce better ads, despite everyone selling ads saying so.

Am I supposed to accept your anecdote as evidence? I mean, on the one hand, we have the massive and continued success of platforms that tailor ads to target audiences, and on the other, we have freeone3000's complaint about a pizza ad not being relevant.

Ad targeting works. That's a fact. You can argue that we should kill ad targeting anyway, that's fine. You can make that case, and I'd disagree. But to flatly deny that ad targeting works? That's adopting a private set of alternative facts, and if you do that, we can't have a conversation.

Have you seen targeted ads that do work? An advertisement for a product you actually want?

Yes. Do you have an argument based on evidence, or at least reason, or are we going to continue like this?

Your arguments are irrelevant because they do not match my experiences. Are you telling me to disbelieve my eyes?

Yes, I participated in a clinical trial I only found though Facebook ads. Since clinical trials are very specific on who they are recruiting, it's a great use of targeted ads.

I also once bought soap that was advertised to me on Facebook. It's my go to brand of soap now.

I also read here on HN someone successfully using Facebook ads to target people planning on moving in their city to find an apartment before it reached the open market. EDIT: here it is, https://medium.com/good-signals/apartment-baiting-with-faceb...

The FHA is a 50 year old law. Facebook isn't getting in trouble for offering targeted advertising, they're getting in trouble for allowing explicit discrimination on the basis of sex and ethnicity. This has nothing to do with correlations.

So it eliminates the financial incentive for massive private surveillance?

Anything that sabotages advertising, especially ad-tech, is a positive thing for the world.

It stops at the product, no? It seems more acceptable to target advertisements for toys than housing.

Should Facebook be liable for allowing an advertiser to target a housing ad at people with preferences for organic food?

Our legal system will not hold Facebook liable for a case like that.

Should?? Like how things ought to be with the world? Facebook shouldn't even know who prefers organic food. There should be nothing tying that preference to any other aspect of people's lives. Facebook should be more like Reddit where people can have whatever accounts they want and can interact on organic food topics without identifying themselves or tying that preference to other interests. Or maybe Facebook should have an explicit opt-in for users to choose what they're okay with being used for ads… … should? everyone should use an adblocker and block Facebook's ads. Also, Facebook should offer a service where you can choose to portray yourself as whatever mix of qualities you want to set for each time you search for something. So anyone should be able to see what the algorithm delivers for a set of characteristics they can adjust at will…

Also, I should stop procrastinating on HN

> Our legal system will not hold Facebook liable for a case like that.

I have to disagree with you on that. I'd put money on targeting a housing advertisement toward someone with a preference for organic food to be ruled discriminatory within five years. It's an unavoidable logical consequence of the logic driving today's complaint.

The problem with your argument is FB doesn’t matter to your point.

The law already exists if that will happen as you say, it will happen no matter what FB does.

So why should FB get to ignore the law unlike thousands of other businesses? Why do they get an exception?

You are misrepresenting my position. Where did I suggest that FB get an exemption from the law? I never said that FB should get special treatment. For clarity, FB should follow the law and prohibit targeting of housing ads based on this list of criteria.

My original point (which has now been downvoted into oblivion, presumably because it raises uncomfortable questions) is that we as an industry need to ask the question "What next?". We know that these regulations will not in fact stop housing discrimination and that more will follow.

It's disappointing that we're unable to have a frank discussion of this subject, since it's important both legally and economically. All the responses in this thread have essentially been that we shouldn't ask "What next?" because the "next" part hasn't happened yet. That's just ostrich-in-sand thinking.

> We know that these regulations will not in fact stop housing discrimination

We don’t know that. I imagine if that was true it would have been proved clearly in the last 40 years. But the law still exists. I’ve not aware of attempts to repeal it.

> and that more will follow.

Again, we’ve had 40 years for this.

So many people in this thread arguing against HUD seem to be trying to argue like this law is something new, like a soda tax, that has very little history or case law. Where the best we can do is hypotheticals. That’s not the case.

> It's disappointing that we're unable to have a frank discussion of this subject

I find it scary that we’re having a discussion. As I see it the story is ‘FB refuses to comply with 40 year old law’, and there are a lot of people arguing they don’t like the idea behind the law so why should they?

Because it’s a 40 year old law. It’s been to the Supreme Court at least once (2015). You don’t get to choose to ignore it because you don’t like it and then get mad when the government tried to punish you for it.

> All the responses in this thread have essentially been that we shouldn't ask "What next?" because the "next" part hasn't happened yet

FHA was 1968. So 50 years, not 40. We have 50 years of ‘next’ to look at.

> That's just ostrich-in-sand thinking.

What is ignoring 50 years of impact and acting like we don’t know how the law would be enforced?

Again? My read of this story was purely as a ‘tech giant decides to ignore he law because they can get away with it’ and had nothing to do with the law in question.

I should have expected HN to see it as ‘poor company bullied by law that theoretically may be problematic based on someone’s beliefs’. I’ve been here far long enough. I should have known. But I didn’t. I looked at the thread. Sigh.

> We don’t know that. I imagine if that was true it would have been proved clearly in the last 40 years.

I don't think it was possible before today's microtarget-based internet advertising to use demographic correlates to substitute for banned targeting of protected groups. That it's now much more feasible than it's ever been in the past to target protected groups without facially making protected group membership part of the filtering process puts anti-discrimination legislation in a new position. The past 50 years of case law isn't particularly relevant to this new situation: in 1968, it wasn't possible to filter ads based on a suite of seemingly-irrelevant characteristics so as to accurately re-create banned targeting.

There is some precedent (e.g., Griggs) for applying a "disparate impact" standard to anti-discrimination measures. My argument is that this reasoning, if followed rigorously, will essentially prohibit ad targeting altogether, since every meaningful ad targeting criterion will end up targeting different arms of protected groups differently.

I've seen no rebuttal of this point, either from you or from anyone else.

> I find it scary that we’re having a discussion

Now that's a terrifying thought. A discussion!

> I don't think it was possible before today's microtarget-based internet advertising to use demographic correlates to substitute for banned targeting of protected groups.

Legal history would prove you wrong. Racist seem to be willing to spend a very large amount of time trying to come up with ways to not have to live next to/rent to certain groups of people. You didn’t need a Pentium 4 to figure out that far fewer black people went to college in the 70s and requiring a college degree would be a pretty good way of filtering them out of your application process.

By now you've posted like 30 comments in this thread and repeatedly broken the site guidelines in multiple ways. Please stop—intense argumentation with sharp elbows is the stuff of flamewars, not intellectually curious discussion.

https://news.ycombinator.com/newsguidelines.html

I don't buy your argument in the slightest. Slippery slopes exist, but the whole point legally is to NOT allow us to go down such slopes. The law defines the (fuzzy) lines. Once it seems to be slipping down a slope, there will be a point where the law (judges) will stop slipping.

If there's an organic restaurant or grocer in a neighborhood, it makes sense to market to those who prefer organic food. There's nothing stopping anyone of any race or background from preferring organic food other than price or education/awareness etc.

Short of laws that ban targeted advertising in general, there will never be a law saying that you can't target people with some consumer preference.

If you tried to avoid black folks by not showing an ad to people who like stereotypical black consumer products, you'd be more likely to get away with it. There would be some burden of proof that you were doing this as a proxy for the protected class.

The core point is that our laws are more likely to have effective loopholes that undermine the actual intention than to have slippery slopes that take the intention and go way too far with it.

No. Because a preference for organic food isn't a protected class.

What if it correlates with protected group membership, like zip code does?

That's a hypothetical question. Does it really correlate with protected group memberships?

> Does it really correlate with protected group memberships?

Yes: "the vast majority of families (73%) who buy organic describe themselves as white" [1]

Now, you might say, "Okay then. Let's prohibit ad targeting by food preference." But the trouble is that there are hundreds of factors that correlate with protected group membership. Are you going to ban them all?

[1] https://www.foodnavigator-usa.com/Article/2015/04/03/Who-buy...

Yes. Disable micro targeting four housing ad listings.

> But the trouble is that there are hundreds of factors that correlate with protected group membership. Are you going to ban them all?

It's a tough question to answer. I believe you were debating me about the use of ZIP code as a proxy in another comment thread so I'll touch on that also.

If it's found that organic food preference is systemically being used as a proxy for a protected class, as ZIP code has been in the past, then it might end up being banned. By following this process could you end up with a list of hundreds of banned factors? In theory yes, but in practice it seems unlikely, just because of how long it would take to study each factor and go through the process of adding it to a banned list. It took a long time for redlining to be made illegal. The article you provided says that organic food preference is gradually starting to mirror the general population, so if current trends continue it eventually won't be useful as a signal for race-based discrimination. Furthermore unlike redlining, there's no vicious circle of "I eat non-organic food -> I can't get a loan to buy organic food -> I have to continue eating non-organic food". By the time you had enough data to decide whether discrimination based on a particular factor constitutes de facto discrimination against a protected class, that correlation might have ceased to exist.