Do Social Media Child Safety and Content Moderation Features Work? | Laura Edelson, PhD

[Kris Perry]: Welcome to Screen Deep, the podcast where we decode young brains and behavior in a digital world. I’m Kris Perry, Executive Director of Children and Screens. Today I’m talking with computer scientist Dr. Laura Edelson, whose research takes us beneath the surface of social media to examine how platforms work. From algorithms that decide what users see, to the design features that keep them engaged, to the safeguards intended to protect children. Laura is an assistant professor at Northeastern University’s Khoury College of Computer Sciences and co-director of Cybersecurity for Democracy. She also served as chief technologist for the antitrust and civil rights divisions of the U.S. Department of Justice. In one recent investigation, Laura and her colleagues put the child safety features of Instagram, Snapchat, TikTok, and YouTube to the test and found that many failed to work as intended or were difficult for young users to access and practice.

So let’s get started, Laura. Welcome to Screen Deep. It’s a pleasure to have you with us.

[Dr. Laura Edelson]: Well thanks for having me. Glad to be here.

[Kris Perry]: The recent social media lawsuit ruling and settlement against Meta and myriad of state legislative solutions might be ushering in a new wave of social media accountability for harms caused to children. To start at the beginning, from your perspective, as a computer scientist and researcher who’s been looking at these platforms and their algorithms and designs for years, do you believe social media use causes harm to young users? And if so, what are the features of social media design that are most harmful?

[Dr. Laura Edelson]: That is the question, isn’t it? I would say that I believe the answer is yes, for some people. Studies that have tried to find a population-level effect haven’t seen that. However, there do appear to be some people who experience, in some cases, very severe harm from some products and some features. The trouble is that we don’t yet know – we haven’t been able to do all the research we would need to do to say definitively, “These users are more at risk for these types of harms.” So I do think that there are some specific features that some user groups appear to be more at risk from, but getting into the details of exactly which users and exactly with which features, that’s the research that’s going on right now.

[Kris Perry]: I appreciate that level of specificity. So can you define what you mean by population effects and why those might hide individual-level harms?

[Dr. Laura Edelson]: Great question. So, let’s think about this in something we understand a little bit better, maybe like cancer. There are some cancers that pose very real risks for certain subpopulations. There are some cancers that really only affect people with certain genetic markers, for example. But if you have those genetic markers, you have a very high, elevated risk of developing that type of cancer. There are certain – like, certain pancreatic cancers work this way, there are others. However, the general population is not at a noticeable risk of developing that cancer because the population is, you know, it’s small, it’s maybe, like, five percent of the population, something like that, that has that genetic marker. So one of the things that’s really hard about social media is, different people appear to be prone to different types of risks. And so maybe all of those risks are real for some subpopulation, but each of those subpopulations for each of those different risks by themselves are small enough that it doesn’t create that type of risk for everyone in the country. Does that make sense?

[Kris Perry]: It does. Are there design features of social media that have any associations with individual harms? And how would you rank those features in terms of impact and developmental concerns for children?

[Dr. Laura Edelson]: I think that there’s a few that we are definitely starting to understand a little bit better. I think the one that pops to mind are beauty filters. You know, so we do know there have been studies that have shown that users who are exposed to beauty filters and use those beauty filters to manipulate images of themselves, that it impacts their perception of themselves. And that it can lead to negative self-ideation. And we know that negative self-ideation is associated with other – it’s one of the risk factors for developing an eating disorder, for example. And it’s associated with other negative mental health consequences. So I think this is one that we are starting to understand the chain of events that can lead to a harm from heavy use of beauty filters. And, you know, we still need to do more research to really fully understand the mechanism there, but that’s one where the research is getting to be stronger. 

Something else that comes to mind that isn’t specifically tied to a design feature, but is a little bit broader around social media use, is some risks to sleep. So for young people in particular – look, sleep is important for everybody, but particularly for kids, there is a very close relationship between getting a good night’s sleep and many other markers of mental health. So kids who use social media overnight, in, you know, in the nighttime hours, they get less sleep. That’s probably not a surprise. And even a really short period of time, even a few weeks of not getting enough sleep can have a really marked impact on a kid’s mental health and well-being.

[Kris Perry]: Great example, and I really appreciated the cancer analogy, too. That really helps me understand the difference between individual and population level effects. 

There’s also a bit of a dose response effect for some of these features, right? As in, a little bit of exposure might be fine, but we’ve got kids using social media for several hours a day and every day. And if all that content is using beauty filters, the effect intensifies. So as you mentioned, the amount of use can displace sleep or cause body dysmorphia or other effects, right? And that’s dependent on how much you’re using the device, these apps?

[Dr. Laura Edelson]: Yes, this is another great question. There does appear to be, in some cases, a dose response effect. And this is another area where we just need a lot more research to understand exactly what that is. And to give you an example of how we are still exploring what some of these different pathways for that dose response might be, there have been studies that have really focused on the number of pickups. They’re the number of checking events, and that’s the kind of dose response they’re looking at. There have been other studies that look more purely at just the total amount of screen time,  and that’s the dose response they’re measuring. 

What’s really hard about all of this and what makes studying this stuff so hard is just genuinely just getting the data. In order to set up an experiment where you would test some of this stuff,  you really need to monitor every single thing that a kid does on their phone. And that is incredibly expensive research to do. Many parents are very understandably hesitant to enroll their kids in a study where we might collect that kind of data. And so there is absolutely research that’s trying to study these things in a granular level, but just getting the data is incredibly hard. And this is why there probably is a dose response. There have been some studies that have shown this, but we just don’t quite yet understand exactly how it works.

[Kris Perry]: I’m really glad you’re bringing up the need for access to the data that the platforms collect in order to make more conclusive statements about what the effects are and what the dose effects are and how many individuals are being affected. And I think it would make you more comfortable as a – someone in academia, to be able to say these things with more certainty if you had access to the data. And that’s a major challenge. It’s one that I hope over these coming months and years, when courts are weighing in and we are regulating these products more carefully, there might be, you know, a nod towards researchers and their need for data. It seems like a really important piece of the puzzle for figuring out how to protect kids a little better. 

There was recently this big settlement with Meta, which I just mentioned, which required some design changes, such as time limits, restrictions on nighttime use. Did the Meta settlement and new requirements for design features actually address the most problematic features for children as supported by evidence in your view?

[Dr. Laura Edelson]: Look, I completely get why everyone is trying to answer this question of, you know, is there harm? Can we – you know, how concretely can we know that there’s harm? Can we prove it in a court of law? There’s another reason why we need more access to data for research, and that’s that, I mean, a lot of the people who run these companies don’t let their own kids use these products. There is a lot of reason to think that for some kids, these products are harmful. And what that means is the actual question we should be working on is how do we make them safer? I completely get why everyone wants to figure out, you know, how big is the pile of liability? That is an important question to answer. 

But I really, really want us to focus on just how do we solve this problem of making these products safer? And the reason that we have to focus on this is that it’s not going to be like flipping a switch. We do not yet understand – I mean, I’m telling you I think there’s harm. I’m pretty sure there’s harm. I don’t know exactly how it works. I know we cannot fix this problem. We cannot start working on the solutions until we really well understand the harms. And then we’re gonna have a whole other big old pile of work we have to do to figure out how to make things safer. How do we, you know – okay, we know this car is – that a lot of people are dying in this car. Are they – like, we need to figure out, like, are they dying because the glass is shattering in a crash? Are they dying because they’re being ejected from the car and they need better seat belts? Are they dying because there’s no crumple zone and the engine is winding up in people’s laps? We don’t know. We just know that a lot of people appear to be experiencing harm. 

And so getting to the question that you actually asked around what the requirements were in the settlement, we don’t know. I think that some of the provisions in the settlement are the kinds of things that are associated with the things we know are harmful, like restrictions on beauty filters. I think that one makes a lot of sense and is backed by the research we do have. But, you know, for other things, a lot of parents are concerned about compulsive use. They are concerned about their kids using social media late into the night. There are some things that appear to be aimed at that, but I – you know what I don’t know? I don’t know if two hours of social media use, or an hour per product, which is – even if the other major social media companies sign on to this agreement, the threshold that they’re going to be agreeing to is an hour per product. That can be, like, four or five hours of social media a day. I don’t know if that’s too much. I’ll tell you as a parent, I wouldn’t let my kid use five hours of social media a day. But again, we don’t know. 

And so I think I really wish that there had been more of a – whether you want to call it a “duty of care” or some kind of ongoing sense that platforms would have to dedicate time and attention and resources to improving the safety of their products and improving access for the kind of safety research that might come up with answers to these questions. I think there are some options for building that more into the settlement, but I think those are the things that I would be more focused on.

[Kris Perry]: Yeah, I mean I think the other challenge with policy change or design changes, you’re hypothesizing that these changes will result in something safer, but unless you study that, unless you collect data before and after those changes, it’s hard to know that you pick the right features, that you change the right products and so the ongoing need for research up front and then post-change is another, like you say, another bucket of work that researchers would probably really like to do so that we could hone in on what the most impactful changes are. And the kind of research you’re talking about takes time. It, you know, harm is happening in real time, right now, every day. And as you mentioned, parents are really concerned, including you. So let me ask you, what do we do right now? How can science move faster to figure out what should be changed? What can families do in the meantime to prevent harm at the individual level?

[Dr. Laura Edelson]: Great set of questions. I’m gonna start with the last one, because I think that’s probably relevant to the most people. I’ve got just a few points of advice that I think are really general.

Thing number one – don’t let your kids have a phone in their room at night. Sleep is genuinely one of the most important things for kids to stay healthy. And we know having phones in a kid’s room at night is very disruptive to sleep. So thing number one, do not let your kids have a phone or access to social media in their rooms at night. It’s very unhealthy. 

Something else that I would say is don’t let your kids lie about their age on social media. Some of my research has found that – let’s say there’s a lot of room for improvement for the social media safety features that do exist for teens, but they’re better than nothing. There really is a difference between a teen account and an adult account on just about every platform. And so give your kids that degree of safety that does exist. 

I would say after that, the most important thing that I can recommend is to have regular conversations with your kids about their social media use. Ask questions about, you know, “What are the good things you saw today? What are the bad things you saw today? How did that make you feel?” One of the conclusions that I am coming to is that a lot of the risks are very specific to certain kids. Certain kids are going to be more prone to seeing a lot of idealized images of other people’s bodies and then feeling bad about themselves. Not every kid is going to experience that. Not every kid is going to struggle to put down the phone when it’s time to come eat a family dinner or to go to bed. And some kids really will. And so I think if you’re a parent, you are in the best position to know that about your kid and to understand what are the risks that they are most susceptible to. And then, you know, not every platform is the same. Different different apps pose risks. And so one of the things you might be able to say, like, “Look, if Instagram is making you feel bad about yourself, maybe you should, like, hop over to YouTube. Maybe that’ll be a place where you can get the funny videos you want without the dose of feeling bad about yourself.” And so I think just having regular conversations with your kids, not having just one conversation and leaving it at that, because this also changes over time. That’s probably the next most important thing to trying to help them navigate what is a constantly changing landscape.

[Kris Perry]: As you noted, platforms have different risks and many already have child safety features for teen accounts. You co-authored a report this year that looked at child safety features available on popular platforms like Instagram, Snapchat, TikTok, and YouTube. What did you find?

[Dr. Laura Edelson]: We found that more than half of them either did not work or were not accessible to kids. We tested safety features for kids and we – I liken it to we bought a bunch of cars, we wrote down which ones said they had airbags and where, and then we crashed them against a wall. We don’t necessarily know that the airbags protect anyone, but we were able to just see do they go off? And so what we found is that in more than half of the cases, the airbags didn’t even go off. So some features, about 10% of the features that we had identified, appear to have been silently deprecated. So they were announced. There were press releases. But then they, at some point, they just stopped working and the company never told anyone that we can tell. So that – yeah, that’s about 10% of the failures.

There was another chunk of failures where the tool, it’s just broken. We can see that it’s active, it exists, but it doesn’t do what the platform says it’s gonna do. So, for example, there is a feature on most platforms that will restrict search terms that the platform has designated as associated with harm. So think, “anorexia,” something like that. If you want to, if you want to search for like, “How do I be anorexic?” Right? That search will get blocked. However, what we found when we tested this feature is that if you search for – instead of searching for “eating disorders,” what will happen is as soon as you get to “eating dis–” it will auto-recommend a misspelling of eating disorders to, like, “eating disorcers.” And then you can just click on that recommended search and you’ll get eating disorder videos. It will, like, instruct you of, you know, how to develop that eating disorder. So we can see that something exists here, but it’s not effective.

And then there was another category of failure where the feature exists. However, it’s buried in the settings. You have to turn it on, it’s buried in the settings, and the user is never told about it. And we’re talking about safety features for kids. We’re talking about tools that are primarily aimed at 13- to 16-year-olds. So that’s just not developmentally appropriate to expect a 13-year-old to navigate through, like, three pages of settings and then select or type in every, like, bullying and hate term that they don’t want to see in their comments. That’s not developmentally appropriate. So that was our other category of failure.

[Kris Perry]: I mean, these are alarming numbers and examples and, again, emphasize the need for transparency from the companies or with independent reports like yours. Can you share any other examples of features that weren’t working that maybe the companies advertise or promote as a solution to some of the problems that you found?

[Dr. Laura Edelson]: I think one of the things that I was most disturbed by is that across every app, every platform we tested, all of the “conduct risk” tools – they were all broken. So these are the tools that are aimed at preventing cyberbullying. And basically the core reason that most of these were broken is the way that they work is they do what sometimes gets called a “soft prompt,” or a prompt to reconsider. Where if you type in something really mean, something really mean, something really mean that I don’t want to say, you will – if you trigger the right combination of words, you will get a little prompt that says – and in some platforms it’ll just say, like, “I don’t know, this sounds really mean. Are you sure you want to send it?” And then you can just say, “Yes, yes I do.” And in other cases, on other apps, you’ll get something that is slightly harder that will say, like, “You know, you can’t say this swear word.” But then in every case we were able to, like, go in and change an “O” to a zero and then post it anyway.

And I think it gets to what is a – you know, are the people who are making these features, are the companies that are making these features, are they treating it like a box-ticking exercise? Or are they thinking about, “What can I do to make this tool actually effective?” And I think that’s where, again, you know, we need metrics of how often these harms happen. We need good public transparency about how effective some of these tools are. And then we need to make them better – to actually improve them, because I think this is something that a lot of parents worry about on social media. You know, “Is my kid gonna get bullied?” And a lot of people, a lot of kids, report getting bullied on social media. And I think there’s a way to think about it of, like, “Well this is just other kids being mean,” but this is being facilitated and often amplified by the underlying technology. And there are things that we can do to make the technology safer. And that’s why we have got to figure out how to make not every single conduct tool fail, because this is something that is hurting people right now, today.

[Kris Perry]: I mean, this is so heartbreaking because we know children in middle childhood or adolescence are so vulnerable to peer approval, peer support, to know that there’s such easy workarounds to some of these conduct safety features, such as replacing a character in a post. I mean, if they put real efforts into these features, it seems they could improve them so easily. I – were there any platforms that you looked at whose safety features actually seemed to be effective?

[Laura Edelson]: So the highest success rate on any platform that we found was 50%, was the best performing platform. That said, there were some bright spots. There were – and I think it is really important to talk about and to pay attention to what was working, because there were some things that were working. And the example I like to give people is what’s called “TikTok for Younger Users.” So if you go to open a new TikTok account, if you sign up with a birthday that is under 13, what the app will do is it won’t say, you know, “I’m sorry, you can’t have an account.” It will put you into an experience called “TikTok for Younger Users.” And in that app experience, there’s no search, there’s no messaging between users, you get a feed –  you get your TikTok feed – you can have that, and it’s a heavily curated feed. It won’t be just anything on TikTok. It’s only content that has clearly gone through content checks and is deemed as safe for younger users. 

So what they are in effect doing is they are not putting guardrails around the things that are dangerous, they’re just taking those features that pose a risk to users – they’re just taking them out of the experience. So what they have done is in effect they have made the default experience the safest experience. I think that is really useful. In our testing we found, when we were in that experience, we couldn’t get to various types of platform prohibited content. That user couldn’t message. That user couldn’t then go search for “eating disorcer” content. And so that was very effective. I think that, and that general pattern of making the default experience the safest one, I think that is a lesson that definitely other companies can learn from. And, you know, when we are figuring out what do we want a safe experience for kids to look like, I think there’s a lot of lessons there.

[Kris Perry]: I mean, the TikTok example you shared underscores something you mentioned earlier, which is on the parent side, making sure your children are entering their actual age when creating accounts so that any safety features for younger users actually take effect. In reality, a large percentage, if not most, of young users are not being fully honest about their age, from what we understand. 

You have some recent work that looked at the system design of different algorithm feeds. Which algorithms did you look at? What can you tell us about the information that algorithms are using to recommend content? And what outcomes are they designed to produce in terms of user behavior?

[Laura Edelson]: I’ve got a paper that came out I guess a little over a year ago that was the result of a long running audit of feed algorithms themselves. A lot of people are very concerned about feed algorithms and what they pump out to us because we have just moved past the era of social media where people are, like, searching and going from one, you know, bulletin board to another. For the most part, most of the content that people see comes to them in their feeds. And so we studied the feed algorithms of TikTok, YouTube, Instagram, Twitter, and actually Reddit to understand what are the common properties of feed algorithms and how do they vary by looking at some of the code that has been made publicly transparent – for Twitter, in particular – some of the patents that these companies have filed about different aspects of their feed algorithms, and then some of the academic work. I should mention, none of this work specifically relates to children, but also, none of these companies have child-specific feed algorithms. They all have like one algorithm and everybody gets effectively the same. So what are these algorithms doing? I think this is such an important thing for people to understand and it’s a little bit hard to get your head around. 

But the thing to remember is that every algorithm – every feed algorithm, from any of these companies – they are trying to maximize how much you, the user, use the platform. Now, exactly what kind of use, that’s going to depend on their monetization strategy. So for platforms that primarily monetize by showing users an ad, those platforms tend to maximize just for the sheer amount of time that you use. Other platforms monetize not just showing you an ad, but having you click through that ad to some other website. Those platforms that tend to monetize click-throughs, they tend to maximize interactivity. So this is why YouTube can feel so very different from Instagram. YouTube primarily sells just ad views, so they want you to be on YouTube for as long as possible. Instagram sells not just that you saw an ad, but that you clicked through that advertiser site and sometimes even that you bought something. So Instagram needs you to be in much more of a mode of clicking the like button, commenting on things, resharing things to other people because that gets you in the mode where you might click on an ad. And obviously this has really big impacts on the kinds of content, but I just think it’s important to remember, this is really all about selecting content that’s going to drive that user behavior that the platform ultimately monetizes.

[Kris Perry]: It’s fascinating how the different monetization incentives for different platforms create different algorithm feed behavior. How are these monetization incentives putting possibly unsafe ads or content in front of children?

[Laura Edelson]: Great question. I think that the problem that we are just perennially facing – I almost said online, but, you know, maybe in society – is that there’s just a lot of scams out there. You know, I think that every time we have some new pivot to a new type of communication technology, very quickly, people who are trying to scam people – honestly, frequently children – into buying some new nutraceutical, or, you know, we have some new miracle weight loss thing, or something that I’ve been quite disturbed by that I have started to see is ads for the new nicotine pouches being advertised on social media with claims that they help boost your focus and boost your mood. And these are things that the kinds of claims that we know kids are really susceptible to. They’re, you know, kids are under a lot of stress these days and who doesn’t want to hear that there’s a magic pill that will make you, you know, feel less stress and feel happier and maybe lose some weight, too? And so, again, this is nothing new, but it is very quickly adapted to new platforms and so there’s been a problem for literally years with these kinds of bogus claims showing up in social media ads because they are usually not subject to the same kind of vetting and scrutiny that ads in other forms of media are. And now, with the rise of short form video platforms like TikTok, like Reels, like Shorts, this has led to the rise of influencer advertising. And so we are absolutely seeing, you know, a lot of ads that don’t even look like ads, that aren’t even disclosed as ads. They are an influencer telling you how they ordered some peptide from China and they lost twenty pounds. Or they, you know, started using these nicotine pouches and they, you know, did great on the SATs. And so I think this is the kind of thing that I really think is one of the more concerning things as a parent.

[Kris Perry]: Absolutely. I mean, these are you know, they’re the harmful product of say, a social media app, combined with an in-real-life harmful product, like nicotine, is really, you know, profoundly worrisome. This is layering harms, and the lack of protection that you outlined earlier with some of the protections being broken, not working effectively. It’s really kind of a perfect storm of things coming at kids without the protections between them and the product, and them having to filter themselves, and how hard that is. 

And, you know, honestly, speaking of scams and unsafe content, how does content moderation actually work on popular platforms and does removing harmful content actually work in terms of preventing users from seeing or engaging with it?

[Laura Edelson]: It could. It doesn’t. So I have done some studies of the effect of content moderation on user experience, and the long story short of the impact of content moderation is that it doesn’t have very much of an impact on user experience. Not no impact, but the effect is small. And the reason why it is small is that in general, content moderation of organic content – I’m not talking about ads here – we studied takedowns of content on Facebook, and what we found is that in general, content takedowns happen very late in the overall life cycle of content.

So on Facebook, which is the only place that I’ve been able to really measure this carefully, content has a fairly short life cycle. It gets most of the views it’s ever going to get in the first 12 hours that it’s up. However, takedowns typically happen after that. Most takedowns happen at about the 24-hour mark. And so what that means is that if you take down the content after most of the people who would ever see it are going to see it, then how can it have much of an effect?

And this, again, gets back to that core issue that people don’t browse a social media platform, they consume what’s in their feed. And so you can think of a content moderation system as being in a bit of a race with the feed algorithm. And so if the feed algorithm is amplifying content really quickly and saying you’re gonna get most of the views you’re gonna get in the first twelve hours – well, if you’ve built your content moderation system that’s supposed to take down things that violate your policies, and that thing takes down most content much later than that, how could it possibly work? Which is not to say that content moderation could not have an impact, but if it was going to have an impact, it would have to be calibrated to actually run ahead of the feed algorithm, not run behind it.

[Kris Perry]: These tools that the platforms promote as ways they’re trying to help solve problems, it sounds good till you start to peel back the layers a little bit, like you’ve done, to say, “Well, okay, exactly how does that work?” And then you find out that it doesn’t really work the way it’s set up. And I mean, is there an answer to content moderation that would do a better job that platforms just aren’t implementing? Or, to put it another way, is effective content moderation that keeps young users from seeing harmful content possible in today’s media environment when speed is the primary answer? Or are there, you know, other methods to content moderation?

[Laura Edelson]: Yeah, there’s lots of ways to do this. I just want to say at the start – what a platform shows a user, I am not the right person to say that. But the important thing to remember is that the core of what any platform for user-generated content that has a feed does, the thing that feed is doing is it’s making choices about what to show a user. And so, you know, I think we talk about content moderation in terms of takedowns. However, there’s this whole other side of it, which is we know that there is more than a billion items of content on YouTube, there’s probably more than a billion items of content on Instagram and on TikTok. You’re only shown a couple hundred items a day in a feed. So a platform is making a massive curatorial judgment in when it chooses what to recommend to you. And so, you know, one, yes, there’s lots of things that platforms can change about how they operate that make their content moderation job easier. However, there – I think the more interesting set of questions to explore are how does a company decide, “Of the billion things I could show a user right now, how do I pick the one hundred that I will show them?” I think it’s a very fair thing to ask, “Hey, if the user in front of you is a child, have some concern about maybe not showing them graphic violence.” Maybe, you know, not showing them other things that we do think are associated with harm in children, but I think no matter what, be really clear and really transparent with users about what you’re doing and what you’re not doing.

One anecdote that I sort of feel the need to share here, because I, to this day, don’t fully understand why this would be the case. When we were doing the auditing of youth accounts, we would set up an account, declare an age as a 13-year-old, and then we would, you know, and then we would set this up to scroll and to use the app. It happened multiple times. I think – I know I saw this multiple times, other testers reported the same thing – where I would be in as a teen account, and in particular, I would see this on on the Instagram teen accounts, and I would get a warning that said, “This thing I’m about to show you has graphic violence. Do you want to see it?” And I would say, “Yes,” and then I’d see, you know, some graphic violence. And I, to this day, find it shocking that in a child account, when you’ve said you’re “not going to recommend graphic content” and then you clearly know that this one is – this isn’t, like, “our classifiers didn’t catch it.”  That’s not what’s going on, because you knew enough to warn me, but then you gave it to me anyway. You know, I – I think this is where there’s some judgment calls that have to be made. Not about what to take down, but what are you gonna choose to show young users? And I think platforms, they are in a very real, in a very literal way  responsible for that, and I hope that they take that responsibility a little bit more seriously.

[Kris Perry]: Yeah, I mean I think what I just heard was you saying a combination of adjustments to the algorithm and better content moderation would get them a lot further than they are right now. And in that right now, they’re not really deploying those tools in a way that is as effective as I bet most parents think they are, that they just sound like they are. Content moderation, I mean, you know, it sounds like that’s what they’re doing, but it isn’t – it isn’t really working as, I’m sure, we think it is.

You had an article in Scientific American last year called “What I Wish Parents Knew About Social Media.” We’ve covered some of your main points from this article already, in terms of algorithms and content moderation. What do you think parents can, and should, do about this information to keep their child safer and more healthy?

[Laura Edelson]: I think one thing that I haven’t yet said, and I think is important to say: you don’t have to let your kids have social media. You don’t. I think that as parents, you know, we want to give our kids everything. And I think that probably for some kids this is completely fine, genuinely poses no risks at all. And they are – they can have a healthy relationship with social media. And maybe you know that that’s your kid. If you think your kid is maybe not that kid, maybe you think your kid gets into social comparisons a lot. If you think your kid – maybe they struggle with putting things down when it’s time to put them down. If you don’t think that Instagram or TikTok or YouTube or any of these things are a fit for your kid, I think it’s okay to say, “No, I’m gonna be the mean mom and tell you you can’t have an Instagram account.”

I am not saying that you should be blind. You know, your kid’s probably gonna try to open an account at school. I do think, though, that again, like, staying in conversation with your kids, making sure that you have a relationship where you can talk openly. You know, if you really think that social media is going to be particularly bad for your kids, you should explain that to them. I do think that, you know, when you talk to kids about what they think about social media, a lot of them are really skeptical because, you know, they see their friends struggling with it or maybe they’ve experienced that they don’t like their relationship with it. And I think sometimes, for some kids, maybe you do have to be the parent.

One thing I want to say is, like, that is not a universal piece of advice. And maybe the universal piece of advice is: talk to your kid and figure out what the right thing is for your kid and your family.

[Kris Perry]: Yeah. I mean, I really love the simple message of, “You don’t have to provide them with access to social media. It’s an option.” And it’s great to step back and just remember something so basic and simple. 

My thanks to Laura Edelson for joining us today on Screen Deep. Laura’s work reminds us of the importance of looking past whether a safety feature exists, to whether it actually works for young people in the real world. Independent research and greater transparency can help parents, policymakers, and the public better understand the systems shaping children’s digital experiences and where stronger protections may be needed. 

You can find more information about Laura’s research along with resources from children’s screens in our show notes. If you enjoyed this episode, please follow Screen Deep wherever you get your podcasts and share this episode with a friend. I’m Kris Perry, thanks for listening, and I’ll see you next time on Screen Deep.