Chris Willis and Lori Sommerfield discuss the new guidance issued by the Department of Housing and Urban Development (HUD) on targeted advertising for housing and housing-related ads.
In this episode of The Consumer Finance Podcast, Chris Willis is joined by Troutman Pepper Partner Lori Sommerfield to discuss the new guidance issued by the Department of Housing and Urban Development (HUD) on targeted advertising for housing and housing-related ads. The conversation delves into the implications of the guidance, which was motivated by HUD's original charge of discrimination against Facebook in 2019 and President Biden's 2023 Executive Order on the Safe, Secure, and Trustworthy Development and Use of AI. They explore how the guidance shifts the focus from disparate treatment to disparate impact, and the challenges advertisers and advertising platforms may face in complying with the new guidelines. The episode concludes with a discussion on the potential for regulatory overreach and the possibility of litigation.
The Consumer Finance Podcast: A Deep Dive into HUD's New Guidance on AI-Driven Targeted Advertising
Host: Chris Willis
Guest: Lori Sommerfield
Date Aired: May 30, 2024
Chris Willis:
Welcome to The Consumer Finance Podcast. I'm Chris Willis, the Co-Leader of Troutman Pepper's Consumer Financial Services Regulatory Practice. I'm glad you've joined us today, because I'm going to be talking with one of my partners, Lori Sommerfield: about HUD's new guidance related to targeted advertising for housing and housing-related ads.
Before we jump into that very interesting topic, let me remind you to visit and subscribe to our blogs, TroutmanPepperFinancialServices.com and ConsumerFinancialServicesLawMonitor.com. Don't forget about all of our other podcasts. The FCRA Focus, all about credit reporting. We have The Crypto Exchange, all about crypto. We have Unauthorized Access, which is our privacy and data security podcast. Finally, Payments Pros, which is all about the payments industry. All of those podcasts are available on all popular podcast platforms.
Speaking of those platforms, if you like this podcast, let us know. Leave us a review on your podcast platform of choice and tell us how we're doing. If you like listening to our podcast and reading our blogs and alerts, don't forget to check out our handy mobile app. It's available for both iOS and Android. Just search for the words Troutman Pepper in your app store. Download it and give it a try. You can listen to all of our podcasts right there in the app, read all of our blogs, see all of our alerts and advisories, and even access a directory of all of our financial services lawyers here at Troutman. So, just give it a try, I think you'll like it.
As I said, we're going to be talking today about HUD's new guidance about targeted advertising using artificial intelligence that was just released recently. I'm going to be discussing that with my partner, Lori Sommerfield, who is a very, very experienced, and knowledgeable fair lending lawyer. Lori, thanks for coming on the podcast today to talk with me about this.
Lori Sommerfield:
Thanks for having me, Chris. Good to be here.
Chris Willis:
So, this is a piece of guidance that just came out from HUD, and just allow me to reminisce for a moment. Because, the issue of targeted advertising being a potential violation of a discrimination law, like the Fair Housing Act, or the Equal Credit Opportunity Act, really made its debut as an issue with the federal regulators with HUD. HUD is the one that started all of this off, because HUD filed a charge of discrimination in 2019, dealing with Facebook, allegedly targeting ads based on the use of protected characteristics under the Fair Housing Act.
For a long time thereafter, HUD was the only federal regulator that had said anything about targeted advertising and what was okay, and what wasn't okay, or mainly what wasn't okay in the charge of discrimination. So now, it's 2024, it's five years later, and a couple of months from when the HUD charge was filed that I've mentioned a moment ago. So, Lori, let's start off by talking about what has HUD come out with and what was the impetus for it?
Lori Sommerfield:
Well, this guidance, I think, is motivated in part from its original charge of discrimination against Facebook in 2019. Just as you said, Chris, I think that HUD has been constantly focused on this issue of targeted advertising, and how it could violate the Fair Housing Act. I think another motivator for HUD in issuing this guidance is that, it was seeking to fulfill its commitment to President Biden's executive order; the safe, secure, and trustworthy development and use of AI, which was issued in October of 2023. That applied to HUD as well as other federal agencies. But that was basically eliciting a pledge from the federal agencies to enforce the federal civil rights laws with regard to new technology like algorithms, and artificial intelligence that are being used in areas like an advertising of housing-related services on digital platforms.
Chris Willis:
That's the background of it. We've got HUD in it in 2019. Then, we have these other background facts. Why don't you tell the audience a little bit about what HUD has said, and the guidance that just came out. In other words, what's not okay, what's a potential violation of the Fair Housing Act dealing with targeted advertising according to HUD?
Lori Sommerfield:
Well, this guidance is really quite fulsome, and that's probably the first that we've seen from HUD since this all began five or so years ago. But HUD's guidance on the applicability of the Fair Housing Act to advertising of housing, credit, and other real estate related transactions on an online platform highlights the potential risks of deploying automated targeting advertisement tools and delivery functions that use algorithms or AI.
HUD asserts that discriminatory advertising can contribute to reinforce or perpetuate residential segregation and other types of harms that can be addressed by the Fair Housing Act. Now, HUD asserts in this guidance that violations of the Fair Housing Act may occur when there's certain ad targeting and delivery functions that unlawfully discriminate on the basis of consumer protected characteristics. Such as, limiting or denying consumers information about housing opportunities, if they don't get to see the same types of ads. Like you and I were talking about, Chris, if you have a different advertisement that goes to a family with children, versus a couple that it's over the age of 62, and is childless. That can set up a dichotomy in terms of targeting, and perhaps unlawfully discrimination on the basis of consumer protecting characteristics.
HUD also argues in this guidance that the Fair Housing Act can be violated when targeting vulnerable consumers for predatory products or services, discouraging, or deterring potential consumers. Advertising different prices or conditions to consumers, and steering home seekers to particular neighborhoods. In some cases, even charging advertisers higher amounts to show ads to some consumers can be a violation of the Fair Housing Act.
Chris Willis:
So, we've got sort of HUD talking about different ways that advertisers for housing or housing related products, including mortgage loans might violate the Fair Housing Act in HUD's view. Does HUD recommend any best practices for advertisers? And if so, what are they telling them to do?
Lori Sommerfield:
Yes, they certainly do. HUD recommended several best practices that in the agency's view, prevent discriminatory uses of AI and advertising. So, they break it into two different categories. There's best practices for advertisers, and then best practices for advertising platforms. For advertisers, HUD recommends using ad platforms that manage the risk of discriminatory delivery of housing-related ads, through audience selection tools, and algorithmic functions. Audience selection tools can be things like categorizing your audience, or engaging in custom audience tools, or using mirror audience tools. That's the type of thing that they're getting at there.
Another recommendation is to follow the ad platform instructions to ensure that advertising related to housing are identified as such, to make sure that they ensure appropriate treatment. HUD also recommends with regard to advertisers that they carefully consider the source and analyze the composition of audience data sets that are used for these custom or mirror audience tools for housing-related ads. Finally, that advertisers monitor outcomes of advertising campaigns for housing-related ads to make sure that they are identifying and mitigating any potentially discriminatory outcomes.
Now, with regard to best practices for ad platforms, HUD has another slate of suggestions that the platform should use. The first is ensuring that housing-related ads are one in a separate process, and that specialized interfaces are designed to avoid discrimination in audience selection and ad delivery. Second, HUD recommends avoiding providing targeted options for housing-related ads that can either directly describe or relate to FHA protected characteristics. Or they can effectively serve as proxies for protected class individuals, either alone or in combination.
Third, this is an interesting recommendation, because I think it'd be very challenging to implement. But HUD recommends conducting regular end-to-end testing of advertising systems to ensure that any discriminatory outcomes are detected, such as by running pairs of ads for equivalent housing opportunities at the same time, and comparing the demographics of the audience to which it's delivered.
Chris Willis:
Let's stop and talk about that for a second, because there's a couple of things I'd like to unpack with you there, Lori. First of all, if I go back to the 2019 charge of discrimination that HUD filed. When I read that charge, and then when I read the subsequent DOJ complaint in the same case, I was struck by the fact that the case seemed to be based on a disparate treatment sort of concept. Where there was explicit use of protected characteristics to target ads for housing. Like black people would get one ad, white people get another ad, something like that. That was the allegation in the DOJ’s complaint, and likewise, in HUD's charge of discrimination. But when I hear a recommendation from HUD about monitoring for outcomes, that sounds like disparate impact to me, not disparate treatment. What do you think about that?
Lori Sommerfield:
I agree with you, Chris. I mean, this is definitely, I think, one of the key takeaways from this guidance is that HUD seems to be shifting from a prior approach of alleging disparate treatment to alleging disparate impact. I think, this particular recommendation clearly speaks to that issue when you think about comparing the outcomes of delivery of equivalent ads to different audience demographics. I think you see this sprinkled throughout the guidance.
Chris Willis:
Okay. So we have this major shift in HUD's discussion of this issue from disparate treatment to disparate impact between 2019 and today? Let's keep going through the recommendations and see what else we see that might be new and interesting.
Lori Sommerfield:
Absolutely. In fact, the next recommendation from HUD speaks to disparate impact very clearly. So, HUD recommends that ad platforms proactively identify and adopt less discriminatory alternatives for AI models and algorithms, including assessing data that's used to train the AI models, and verifying that the technologies measure lawful attributes that predict valid outcomes. Clearly, evaluating less discriminatory alternatives, or LDA as we call it in the biz, is clearly an indicator of a disparate impact legal theory here.
Chris Willis:
Okay. Not only do we have the idea of disparate impact, but we have like the very leading edge of what we're seeing the CFPB talk about. Not super publicly, but at least talk about some, and to talk about in supervision. Which is, they view it as kind of a requirement to look for less discriminatory alternatives for underwriting models for credit products. Now, you have HUD taking that concept and applying it to an advertising model under the Fair Housing Act, which also strikes me as a very large step forward beyond where we had been before.
Lori Sommerfield:
I completely agree with you, Chris. And it seems like now, HUD is on the bleeding edge of what should be required at this point for advertisers and advertising platforms. If I can just quickly mention the last three recommendations HUD had for ad platforms. Let me just wrap up this discussion. But the other three recommendations were that ad platform should ensure that algorithms are similarly predictive across protected class groups, and make any adjustment to correct for any disparities, and predictiveness, or direct the algorithm to develop additional information that can enhance predictiveness for certain protected class groups.
Chris Willis:
Well, hold on a second. The advertising platform is supposed to measure predictive power by protected class. If the algorithm is better with one than the other, it's supposed to fix it by what? Taking into account protected class, and the algorithm, and like fudging it.
Lori Sommerfield:
That is not explained in the recommendation. But to your point, it sounds like that would be very difficult to comply with.
Chris Willis:
Well, it seems like you'd have to actually explicitly take protected class into account in order to achieve what HUD is saying as a requirement, which again, taking protected class into account is exactly what the law prohibits.
Lori Sommerfield:
Right. I think that that particular recommendation is going to require some clarification from HUD, about how ad platforms can comply with it without running afoul of the Fair Housing Act.
Chris Willis:
Yes. I apologize for interrupting you there. I just sort of fell off my chair a little bit there and I was having to recover. Let's talk about the rest of the best practices.
Lori Sommerfield:
Okay. Just two more. Ad platforms are also supposed to ensure that ad delivery systems are not resulting in differential charges on the basis of protecting characteristics, or charging more to an advertiser to deliver ads to a nondiscriminatory audience.
Chris Willis:
Okay.
Lori Sommerfield:
Okay. Then finally, ad platforms are suggested that they document, retain, and publicly release in depth information about their ad-targeting functions, as well as their internal audits about the model.
Chris Willis:
Oh, sure. Well, there's a huge historical precedent of that happening. I'm sure that's going to occur like immediately, right?
Lori Sommerfield:
Highly unlikely.
Chris Willis:
By the way, going back to the charging issue, like you can see if you don't think about it too hard that, "Oh, of course, it makes sense." I wouldn't want to charge an advertiser more to advertise to Hispanic people versus white people or something like that. That would make sense. But if you're applying a disparate impact analysis to it, and let's say that you get charged more for someone who's manifested a recent interest in a mortgage loan than someone who hasn't manifested a recent interest in a mortgage loan. That seems very appropriate from a business justification standpoint. But what if there's not equality among protected classes in that? So, does that equate to charging someone more on a discriminatory basis if the population that's targeted for a legitimate business reason ends up not being equal? This is why I think applying disparate impact to advertising is so problematic.
Lori Sommerfield:
I could not agree with you more, Chris.
Chris Willis:
Okay. So, it seems to me, as you can tell by my questions, that the recommendations that HUD is making seem to go well beyond what I think would normally be expected, both for advertisers, and for ad platforms themselves.
Lori Sommerfield:
Agree.
Chris Willis:
I guess we've previewed some of this. But I mean, what challenges do we foresee for compliance with these HUD best practices in that regard?
Lori Sommerfield:
Well, I think the testing aspect, the recommendation that ad platforms test end to end, that's going to be very challenging, as we've discussed. The recommendation to run pairs of ads for equivalent nondiscriminatory outcomes when you're using the same ad with the same housing opportunity. That is just going to layer so many requirements like that on ad platforms that will be very difficult for them to comply with.
But for advertisers too, advertisers generally have to rely on limited information about consumers when they are advertising. They don't know who's actually going to be seeing the ad necessarily, because they haven't submitted an application, they haven't provided any personal information. Advertisers also do not have access to that much information about the potential audience that may view the ad. So, I also think that another challenge is that HUDs guidance seems to imply a duty for housing providers to ensure that advertisements are seen equally across members of all protected classes. That's likely going to be very difficult to achieve compliance with under the Fair Housing Act.
Chris Willis:
Right. Because the thing is, there's only one advertising platform in the country right now that has a variance reduction system. That's the one that's huddled with the Department of Justice. How does that advertisers supposed to get assurance from every other single one in the country that they have something that works like that? Which seems to be what HUD is demanding.
Lori Sommerfield:
Exactly. I think this all remains to be seen. This guidance, as you pointed out, just simply goes much farther than any other guidance we've ever seen, issued by either HUD or any other federal regulator.
Chris Willis:
Okay. Well, I guess we'll have to stay tuned and see what happens, and see if the guidance is followed by enforcement behavior from the regulators that actually have enforcement authority, like for example, the Department of Justice, or if we get any written guidance from any of those other regulators that sort of goes down this line? Or if HUD's going to stay sort of an outlier in this regard? We just have to wait and see now, I guess.
Lori Sommerfield:
Exactly. It'll also be interesting to see if it is challenged by any advertisers or advertising platforms that find that they cannot comply with it.
Chris Willis:
Yes, and that makes sense. There's quite a lot of litigation over industry, being dissatisfied with what they perceive as regulatory overreach these days. So, very well could result in litigation, as many other regulatory announcements have. Well, Lori, thank you very much for joining me to talk about this with our audience. This is a really important and fascinating issue and your insight of it is incredibly helpful. Thank you for being on today.
Lori Sommerfield:
Thank you for having me, Chris.
Chris Willis:
Thanks for our audience for tuning in today too. Don't forget to visit and subscribe to our blogs, TroutmanPepperFinancialServices.com and ConsumerFinancialServicesLawMonitor.com. While you're at it, why not head over to Troutman.com and add yourself to our consumer financial services email list. That will allow us to send you copies of the alerts and advisories we send out from time to time, as well as send you invitations to our industry-only webinars that we also have pretty frequently. Don't forget about our mobile app. It's available for iOS and Android, just type in Troutman Pepper in your app store. Of course, stay tuned for a great new episode of this podcast every Thursday afternoon. Thank you all for listening.
Copyright, Troutman Pepper Hamilton Sanders LLP. These recorded materials are designed for educational purposes only. This podcast is not legal advice and does not create an attorney-client relationship. The views and opinions expressed in this podcast are solely those of the individual participants. Troutman Pepper does not make any representations or warranties, express or implied, regarding the contents of this podcast. Information on previous case results does not guarantee a similar future result. Users of this podcast may save and use the podcast only for personal or other non-commercial, educational purposes. No other use, including, without limitation, reproduction, retransmission or editing of this podcast may be made without the prior written permission of Troutman Pepper. If you have any questions, please contact us at troutman.com.