New Mexico Judge Appoints Himself Meta’s New Product Manager

from the where’s-his-performance-review? dept

Earlier this year, following the initial verdict in the state of New Mexico’s case against Meta (arguing that Meta caused harm to children by doing things like encrypting messages), we noted that even if you believe that Meta is a terrible company (as I do), that Mark Zuckerberg cannot be trusted (ditto), and that the company has prioritized profits over safety (yup), we should be seriously concerned about the verdict. Among the many problems with the verdict is that it’s a judge and jury taking on the role of determining what they feel is how internet apps can and should work, despite having no knowledge or expertise regarding how internet products can and should work — or how trust and safety tradeoffs actually work.

The second part of that trial, focused on remedies beyond the monetary judgment the jury already awarded the state, concluded last week. Almost all of the headlines are talking about the additional $567 million the judge ordered the company to pay above and beyond the $375 million in civil penalties the jury already awarded. Those funds are supposed to be used to help pay for the supposed “harms” of Meta’s properties to New Mexico kids. But that is probably the least interesting part of the ruling.

What’s way scarier is that the judge then makes a bunch of design decisions, telling Meta how its products need to work. Admittedly, many of the changes are ones Meta itself suggested (some of which it has already made). But this is exactly the kind of thing that I was worried about when the first verdict came down. Once we’re in a world where a judge gets to make product decisions, bad things are going to happen. Judges are not product designers. Judges are not trust & safety experts. Judges are not child safety experts. They do not understand how all these things work together. They do not understand the tradeoffs of their decisions.

Judges simply shouldn’t be in the business of determining the proper user interface for software. Indeed, I’d deem it to be an example of compelled speech and (in cases of banning certain features) suppression of speech.

In the ruling, the judge now says that Meta will need to limit the amount of time children can spend on the app (which means mandating age verification of every user), that the number of “likes” on photos on kids’ accounts must be hidden, and then some compelled “warnings” about the supposed harms of social media. All of this is problematic.

Again, even if you hate Facebook and think it’s harmful and you trust this judge, do you trust all judges out there to get this right? Do you think that some of the Trump appointed judges should be able to decide what features certain apps should have? We’ve already seen some judges who think that any trust & safety/content moderation operation is harmful. And we’ve seen judges on the flip side who think that not taking down certain speech is equally harmful. Letting individual judges determine how products are designed is a disaster waiting to happen.

As we’ve discussed for years, there is scant evidence that social media (and Facebook’s apps in particular, as they’re the most studied) are inherently harmful to kids. Many, many, many researchers have tried to find such evidence and tend to come up empty. It would be one thing for the judge in this case, Bryan Biedscheid, to just ignore all of that, but he doesn’t. He more or less admits it, highlighting the various claims from both sides in the case… and then just declares which side he thinks is more credible. But a different judge might find otherwise. Which is a big part of the problem here.

If all the science agreed, this would, perhaps, be an easier call. But the vast majority of the science disagrees with Biedscheid’s feelings here.

He starts out by citing the Surgeon General’s report from 2023, but reads it to say a lot more than it actually does:

In 2023, the U.S. Surgeon General issued a public advisory titled “Social Media and Youth Mental Health” (“2023 Advisory”) which “calls attention to the growing concerns about the effects of social media on youth mental health.” [Pl. Ex. 03193 at 3] As the 2023 Advisory notes, “[a] Surgeon General’s Advisory is a public statement that calls the American people’s attention to an urgent public health issue,” and such advisories “are reserved for significant public health challenges that require the nation’s immediate awareness and action.” [Pl. Ex. 03193 at 3] The 2023 Advisory states that “[u]p to 95% of youth ages 13-17 report using a social media platform, with more than a third saying they use social media ‘almost constantly.’” [Pl. Ex. 03193 at 4] The 2023 Advisory also notes that “nearly 40% of children ages 8-12 use social media.” [Pl. Ex. 03193 at 4]

Importantly, the 2023 Advisory reports that features designed to maximize engagement can harm children by encouraging problematic use and behaviors. The 2023 Advisory provides that:

[e]xcessive and problematic use of social media can harm children and adolescents by disrupting important healthy behaviors. Social media platforms are often designed to maximize user engagement, which has the potential to encourage excessive use and behavioral dysregulation. Push notifications, autoplay, infinite scroll, quantifying and displaying popularity (i.e., “likes”), and algorithms that leverage user data to serve content recommendations are some examples of these features that maximize engagement.

But that proves nothing. Note even the hedging language here. It says that excessive or problematic use can harm children… but doesn’t say how often that occurs. It says that those features have “the potential to encourage excessive use” but makes no findings on whether they actually do or how frequently they do.

Indeed, as we keep pointing out, where the evidence actually takes us is that a very small percentage of kids cannot handle unsupervised, unlimited social media. Most of the studies appear to put it at less than 5%. And, much of the research seems to suggest that any causal connection is in the other direction. That kids who are not getting the mental health support they need then turn to social media and use it excessively, rather than the social media causing the problems.

Also, the judge completely leaves out that the same Surgeon General report talks about how incredibly helpful social media is for many kids. He does mention elsewhere other studies showing that social media has benefits… but then forgets all about it.

The evidence in this case demonstrates that social media has many benefits, such as helping people connect with friends and family. [See, e.g., 2-27-26 Tr. 7225:11-14 (Cain); 2-19-26 Tr. 4904:1-11 (Coyle); Def. Ex. 01045 at 93] It also helps members of marginalized communities form and build online communities based on connections and shared interests, especially when members of those communities might not have access to supportive or even safe communities offline. [2-12-26 Tr. 2452:22-25 (Boyle); 3-11-26 Tr. 10882:22-10883:16 (Otaru)] It is also a source of news, educational content, and entertainment. [Def. Ex. 01045 at 93] Meta’s services in particular help small businesses grow by allowing them to reach more customers.

In other words, the issue is not about social media inherently. And any solution across the board is a terrible idea. Instead, we should be making efforts to identify which kids are actually at risk and helping them. Not just universally declaring which features are good and which are bad.

Unfortunately, that’s what the judge does here, effectively appointing himself the product manager for Meta’s social media products and saying that despite the conflicting evidence, he’s convinced that Meta’s products are designed to be harmful to children:

Based on the evidence in the record, the Court finds that Meta implemented platform features that were designed to optimize engagement, and that these features were and are harmful to teenagers. [See, e.g. 2-17-26 Tr. 3573:15-17, 3581:15-3582:16 (Narayanan) (outlining how Meta’s algorithm uses engagement-maximizing recommendation algorithms); 2-23-26 Tr. 5468:22-5470:15 (Lembke) (describing how children are “uniquely vulnerable” to harm on social media because their brains are still developing)]

But… optimizing engagement is what every cultural product does. Many novels I read end each chapter on a cliffhanger, and sometimes that gets me to stay up late at night as I need to know what happens next. That’s harmful to my sleep. And it’s a design decision from the author/publisher to end chapters like that. But I think we all recognize that it would be a massive First Amendment problem if we told publishers that they couldn’t end chapters on cliffhangers, saying that they “optimize engagement” and therefore are “harmful.”

Basically, the court decides that the experts the state put on the stand were more credible than the experts Meta put on the stand. But just the fact that there’s so little agreement about what the science actually says here, and no study that has shown causality, should give the judge pause. Instead, because one witness claimed a causal relationship, he accepts that as fact.

It would be nice if the judge showed a bit of humility here, but instead seems to assume that based on a few witnesses at the trial he can redesign Facebook and Instagram.

The judge, who had previously rejected Meta’s Section 230 defense, admits that he can’t do anything regarding actual content on the app, because that would likely violate both 230 and the First Amendment:

This decision does not seek to close or demolish Meta’s platforms, enjoin specific advertising or content (which the Court is also mindful is protected by the First Amendment and 47 U.S.C. Section 230 of the federal Communications Decency Act (CDA)), or require Meta to cure causally disconnected harms. Rather, this decision seeks to address existing harms created by Meta’s platforms and to prevent future harm to children and future burdens on New Mexico, in general, that would otherwise be caused by Meta’s platforms. It is this Court’s conclusion that Meta is a cause of and has substantially contributed to a public nuisance in New Mexico, and is required to abate that public nuisance to the extent of its contributions.

But Judge Biedscheid still appoints himself the new product manager for safety for Meta’s products for other features. He establishes that Meta’s social media product features represent a “public nuisance” to kids in New Mexico, and they must abate the harm of that public nuisance. First, it tells Meta that it needs to implement age verification, though it admits that it’s somewhat limited by federal law (COPPA) in how much it can require:

Age verification is the key to making Meta’s platforms safe for adolescents because there must be substantial certainty about whether a user is: (a) over 13 years of age, thus old enough to use the platforms; and, (b) over 18 years of age, so that adolescent protective restrictions on use should be removed and restrictions on the ability to connect with adolescents must be put in place. Without substantial certainty, there will be adults communicating with adolescents by claiming a false younger age and tweens communicating with teens and adults by claiming a false older age.

But… most experts don’t think that’s true. Indeed, many experts have made it clear that age verification introduces all sorts of new risks and dangers. The letter linked there (from 438 experts) was even brought up during this phase of the trial, but was dismissed as being by experts who were out of touch by New Mexico’s experts (and, apparently, the judge).

And it’s this kind of thing that feels so problematic about this. The science isn’t even remotely settled here, and there are credible experts warning that these solutions will actually do way more harm than good for children’s safety. Shouldn’t that be important too?

What if these changes demanded by the judge actually do more harm?

Then what? Do we get to sue the state of New Mexico for being a public nuisance? Or sue Judge Biedscheid? Nowhere in this order is any sort of humility or acknowledgement that he has no idea the actual impact of these changes. There is no plan to go back and look and check to see if they are working. Or if they’re causing more harm. If Meta makes these changes and the rate of youth depression or suicide increases next year, then what? Whom do we get to sue? Who “abates” that public nuisance?

And then there are other features that Product Manager Biedscheid decides Meta can no longer offer to kids (as determined by the age verification tools he’s requiring them to use):

Meta shall eliminate push notifications on its platforms for known or estimated accounts belonging to users under 18 years of age: (a) from 10:00 PM to 7:00 AM (i.e., hours when the vast majority of children are sleeping or should be asleep) on all days; and, (b) from 8:00 AM to 3:00 PM (i.e., typical school hours) during the academic year, excluding weekends. Notwithstanding the foregoing, Meta may make exclusions to this rule for messaging from connected users and for urgent, targeted messages, such as for security or hazard alerts

Meta shall hide, as a default setting for all accounts where the user is under 18 years of age, all “like counts” that Meta appends to content. Meta may only allow an override of the default setting when a parent or guardian gives the user under 18 years of age their permission to change the default setting.

Meta shall implement a mandatory usage time limit for accounts belonging to users under 18 years of age. Meta shall restrict the usage of all such users to not more than 90 hours of use per month cumulatively across Facebook and Instagram.

Do we know if any of this will help kids? ¯_(ツ)_/¯

Will the court go back and revisit this if this magically doesn’t help kids mental health struggles? ¯_(ツ)_/¯

The court also mandates a bunch of compelled speech that the Zauderer standard says it has no business compelling. But the judge doesn’t seem to much care:

Meta shall provide information screens on Facebook and Instagram to be displayed to all new users under 18 years of age once a day that explain one or more of the following: (a) an aspect of safe platform use best practices; (b) available tools to address inappropriate content and behavior; (c) the concept of rabbit holing and ways to reset the algorithm to address the issue; and, (d) tools to avoid problematic use, such as nudges, self-imposed limits and other tools. These screens shall be displayed for the first 30 days that a new user under 18 years of age joins a platform, similar to Meta’s presentation of new users under 18 years of age with Pristine Pool content. The screens shall be submitted to the State for its review, possible edits, and approval.

Meta shall fund, design, and implement an educational campaign in New Mexico, prepared in collaboration with and with the approval of the State, that publicizes: (a) risks associated with platform use; (b) safety tools and best practices to address those risks; (c) parental controls; (d) problematic content and behavior reporting tools; (d) risks associated with online bullying and features that help victims address it; and, (e) reporting mechanisms for accounts belonging to users under 13 years of age. This campaign shall also provide schools with ready-made materials, also prepared in collaboration with and with the approval of the State, that set forth information on these topics to be distributed as desired by the State.

It’s entirely possible that those are good ideas. But we should all be worried about judges ordering internet companies what they must tell their users. Remember, we’ve seen other attempts to do this — such as the law in Texas that tried to force adult content sites to warn people that porn was dangerous. And those were thrown out as unconstitutional.

But how do you distinguish these warnings from the anti-porn disclosure warnings pushed by the far right? The point is that the state isn’t supposed to be in the business of requiring warnings on inherently speech related products. Especially when there is no agreed upon scientific basis for the claims.

About the only good thing in the ruling is that the judge neglects to take the state up on its worst suggestion: banning end-to-end encryption in WhatsApp or Facebook. But even that is only partially good, because he orders Meta not to turn end-to-end encryption back on for Instagram. As you’ll recall, back in May, Instagram turned off end-to-end encryption, claiming that no one was using it. Privacy advocates have asked the company to turn it back on, but if this ruling stands, that won’t be allowed. The judge’s rationale is basically that (1) WhatsApp doesn’t have the other “addictive” features he dislikes and (2) kids don’t use Facebook anyway, so only Instagram is the problem:

The Court does not adopt Plaintiff’s requests for a prohibition on end-to-end encryption (“E2EE”) for the following reasons. Meta has already ceased offering E2EE on its Instagram platform, and the Court orders that such cessation of E2EE on Instagram remain in place during the Abatement Period. Regarding Facebook, the Court does not agree with ending E2EE on this platform. Facebook has few adolescent users in New Mexico, and the Court only has jurisdiction over New Mexico accounts. Therefore, in light of the benefits E2EE may offer in many other markets outside of New Mexico, and the limited use of E2EE by adolescents on Facebook in New Mexico, the Court does not impose a prohibition on E2EE in relation to Facebook.

But, again, on what scientific basis is this decision made? It all feels like vibes.

It’s also weird because a few paragraphs later, the judge admits that the benefits of end-to-end encryption outweigh the harms… but only on WhatsApp? On Instagram it’s the opposite? Why?

To the extent that WhatsApp’s E2EE poses a risk to adolescents, that risk is outweighed by the benefits that E2EE offers to address privacy concerns of New Mexicans and other populations.

He also admits that while he might like to force Meta to turn off autoplay, infinite scroll, and content recommendations, he (correctly) worries that those (since they’re so directly tied to third party content and editorial decision making) could implicate both Section 230 and the First Amendment:

Of the above-mentioned design features, autoplay, infinite scroll and algorithmic content recommendations are most closely tied to content presentation because they directly impact the manner in which users are presented with third-party postings and advertising. As a result, autoplay, infinite scroll, and algorithmic content are features with clear Section 230 and First Amendment implications.

The evidence and argument at trial also showed that autoplay, infinite scroll, and algorithmic content recommendations are used widely in the industry. Therefore, in light of the broader market in which Meta operates and given the absence of Meta’s competitors in this litigation, restrictions imposed on Meta’s offering of the aforementioned features could harm the viability of Meta and its platforms.

In contrast, push notifications and “like counts” are least connected with platform content. Indeed, push notifications occur in many instances even when the user is not directly interacting with the “pushing” application on their device. Additionally, “like counts” are merely a feature created and offered by Meta to principally track and motivate user feedback, all without altering the underlying published content.

As with Age Assurance, while the Court agrees with the State that autoplay, infinite scroll, “like counts,” and algorithmic recommendations combine to facilitate addictive or problematic behaviors in adolescent users, the Court does not see an equitable abatement method: (a) that impacts only Meta, rather than imposing restrictions industry-wide; and, (b) that respects the protections of the First Amendment and Section 230.

The simple fact is that all of this is one judge issuing orders regarding how a massive company with billions of users needs to be designed. And, yes, he’s trying to keep the platform safe but he’s making these decisions based on his feelings about what will work, and he’s no expert. And there’s nothing in there considering what happens if these changes actually make things worse, as they absolutely could do.

Within Meta, I’m quite sure they measure every little change. They check to see what works and what doesn’t. And when things don’t work, they adjust. Frequently. They have KPIs to meet and regular reviews. If something goes wrong they roll stuff back re-evaluate. They run A/B tests. They explore each decision, and they have to continually justify the decisions they make.

As their new product manager, Judge Biedscheid has none of that. He gets to declare from on high what changes Meta must make to their product, and… that’s it. The design decisions are his. But the responsibility isn’t. If teen mental health in New Mexico doesn’t improve — or if it gets worse — there’s no plan to review. There’s no rolling it back. There’s no A/B testing. Judge Biedscheid is off on another case.

Filed Under: age verification, bryan biedscheid, kid safety, new mexico, notifications, product design, trust & safety

Companies: meta