Social media
and online harms
How can we protect young
and vulnerable people?
The line between online and offline is increasingly fuzzy. “I used to distinguish between online friendships and offline,” says Dr Nejra van Zalk. “People I know offline, some I know on and offline, and people I only meet face to face.”
Now, though, it’s very different. “Now, I’d say, the majority of our friendships are maintained through online platforms. Either through WhatsApp and other asynchronous interactions, or social platforms like Facebook. The vast majority of young people have social interactions which bleed into online, and interact with people they’ve never even met.”
[Arguments on both sides] are really un-nuanced. There are positive things and negative things.
Van Zalk is a lecturer and researcher in psychology and human factors at Imperial College London’s Dyson School of Design Engineering. The Department for Culture, Media and Sport (DCMS) has begun a consultation ahead of a new white paper on online harms, with the eventual plan of creating an online regulator, and van Zalk and her colleagues at The Forum have begun a response to that consultation, including hosting an online discussion.
Dr Nejra van Zalk is a design psychologist at Imperial's Dyson School of Design Engineering
Dr Nejra van Zalk is a design psychologist at Imperial's Dyson School of Design Engineering
The digital world has changed rapidly, and a lot of people – especially parents – are worried about their children, now that so much of their lives are lived online. There’s a lot of fearmongering and Manichaean black-and-white thinking, says van Zalk, but the real picture is complex.
“I notice a lot of older people writing about this,” she says. “People saying that online interactions are messing children up, turning them into morons.” She points to recent popular books about the “iGen” generation, suggesting that smartphones have damaged young people.
“It’s large-scale scaremongering,” she says. But conversely, “there are fringe voices saying it’ll be a digital utopia, with free information.”
Both of these views, she says, “are really, really un-nuanced. There are positive things and negative things.” Instead, we need to look more closely at what is good and bad about our online lives.
Tech companies are incentivised to get us to spend more of our life on their sites, and that can come at a cost to our physical and mental health, in various ways; but on the other hand, social media links us closer together. Teasing out the effects in detail, positive and negative, is important.
The trouble is that research into online harms is hard; it is hard to tease out causality, and to get good metrics. Online platforms change quickly, making research rapidly out of date. It’s also hard to define what the harms are; “online” is a big place, and talking about “online harms” as one thing means lumping things as disparate as eating disorders, radicalisation, and filter bubbles under one umbrella. Even the measures we use to study it are lacking nuance, says van Zalk.
The need for a more nuanced understanding of the impacts of
digital technology
You might think, if you read popular science books and the press, that there is good, solid evidence that young people are being harmed by the online environment. But, van Zalk said at The Forum’s online panel event in June, it’s not as straightforward as that.
“Sherry Turkel [a professor of social psychology at the Massachusetts Institute of Technology] will tell us that online social interactions are creating social distances, and dwindling attention spans,” she said.
Screen time is a very blunt tool. How long a screen is on doesn’t tell you anything about whether that time is spent well or not.
“Jean Twenge [a San Diego State University psychology professor] wrote a book about the iGen generation, saying that we’re all turning into morons. The vast majority of reports veer towards these negative aspects.”
But in fact a lot of the research is extremely weak. “The academic research is really scattered; it’s mainly cross-sectional studies that fail to capture the complexity of online interactions,” she said. She and colleagues have written a book, Online Peer Engagement: Positive and Negative Aspects of Online Social Interaction, reviewing that evidence and trying to create “a more nuanced” understanding of the impacts of digital technology.
When we discuss online behaviour, most people think in terms of “screen time”, and lots of research looks at exactly that metric. “But screen time is a very stupid and blunt tool,” van Zalk told attendees of The Forum. “How long a screen is on doesn’t tell you anything about whether that time is spent well or not.
“I spend about eight hours in front of a screen [a day], but the majority of those hours I’m educating myself, I’m reading.” Children skyping with their grandparents and someone scrolling endlessly through their news feed or Twitter timeline – or someone watching internet porn or violent extremist videos – are all “screen time”, but they’re very different things.
More than that, she says, most research also relies on self-reported use – people remembering how much they used their phone or other device. Recent research suggests that self-report is highly unreliable. “Even adults report it wrongly,” van Zalk says.
It’s one of my nightmares. Is anxiety a stronger predictor of spending time online, or vice versa? Or is it bidirectional?
There are more problems. It’s very hard to tease out causality – if people who use social media more tend to be depressed, is that because social media makes them depressed, or because being depressed means they spend more time looking at their phone?
Most studies into online harms are cross-sectional – that is, they take a snapshot of the population at a single time. They might, for instance, ask 10,000 people about their phone habits and about their anxiety levels and see whether the two are linked. But while that can show a correlation, it can never show a causal relationship.
The best way to show causality is to do a randomised controlled trial – taking a group of people and deliberately exposing half of them to whatever you’re trying to study. But it might be unethical (and impractical) to do that with online harms.
Dr van Zalk says screen time is not a good metric for studying online behaviour
Dr van Zalk says screen time is not a good metric for studying online behaviour
“It’s one of my nightmares,” says van Zalk. “Is anxiety a stronger predictor of spending time online, or vice versa? Or is it bidirectional – the more anxious, the more online, the more anxious and so on? It’s always going to be a nightmare. I often wonder why I didn’t study glaucoma in mice or something.”
There are ways of establishing causality without a controlled trial. Longitudinal studies, looking at a population over a prolonged period, is one; if you notice that your purported effect appears before your purported cause, then that strongly hints that the causal relationship you imagined is the wrong one. “I primarily do longitudinal studies,” says van Zalk.
Dr Dasha Nicholls, a Clinical Reader in Child and Adolescent Psychiatry at Imperial’s Department of Brain Sciences who specialises in eating disorders, points to a study by the Harvard psychiatrist Anne Becker, which looked at the impact of Western television on Fijian girls’ body image.
She calls it a “naturalistic experiment” in that the rise of television in the late 1990s changed Fijian society; the study found that rates of eating disorders rose after television became commonplace. But even then, we can’t be sure about causality – there may have been many things that changed at around that time.
Dr Dasha Nicholls is a child and adolescent psychiatrist in the Department of Brain Sciences at Imperial
Dr Dasha Nicholls is a child and adolescent psychiatrist in the Department of Brain Sciences at Imperial
To make matters worse, the digital world changes fast. “Researchers are always playing catchup,” says van Zalk. “By the time your study on Facebook is released, Instagram is the popular thing. Or now they’re moving towards Tiktok. What’s a popular platform one year may not be the next year.”
Nicholls agrees: “It changes so rapidly. Our data on cyberbullying is from back when people were using Facebook. No one uses Facebook now. It’s very hard to do the slow, methodological research that normally leads to these answers.”
Nonetheless, this research is important.
There are real harms that young people, especially, may face in the online environment, from the prosaic to the extreme.
Speaking at The Forum’s online panel discussion, van Zalk said young people spend “an average of 135 minutes a day online, and that’s likely an underestimate”, and that according to a recent UK government report, they face dangers from being online including “radicalisation, unwanted sexual attention, pornography, cyberbullying, trolling and commercial risks”.
Eating disorders can be exacerbated by online interactions of various kinds.
At one end are the sort of ordinary experiences that we all have: the aimless flicking through our phone, waiting for the one tweet or post that will tell us to log off.
“There’s really interesting research on young women showing that women who report just scrolling through news feeds with no interaction whatsoever,” says van Zalk. “That’s related to higher levels of psychopathology [mental health issues].”
Then there are more obviously worrying things. “My specialism is eating disorders,” says Nicholls. “They can be exacerbated by online interactions of various kinds.” The most extreme, she says, are the “pro-ANA websites”, which glamourise and encourage anorexia and bulimia.
“But more widely available, and pernicious, stuff is the endless invitations to compare yourself to others and be judged on the basis of your appearance,” she says. Young women seem especially prone to these problems.
Cyberbullying, although not as widespread as some of the most alarmist claims, affects many children. “With cyberbullying, as opposed to traditional bullying, you can be bullied in your bed at night,” says Nicholls. “You’re not safe at home.”
And, she says, the ease with which you can bully someone – the fact that it is easier to send someone a message saying something cruel than it is to say it to their face – lowers the threshold for the sort of person who does the bullying. She took part in one study “on the personality profiles of people who become cyberbullies and how they differ from traditional bullies.”
It wasn’t time online itself [that predicted mental health], but whether it stopped people getting enough sleep or exercise.
“What we found [was that] the availability of online platforms made it easier, so the personality profiles look more normal - traditional bullies have personality profiles similar to those of aggression and contact disorders.” So they might score highly on personality measures such as narcissism or callous-unemotional traits, or have low self-esteem.
Then there are the more extreme, but rare, things. “We have very real reasons to be worried about online grooming, and different forms of radicalisation, and extremist content online, child trafficking, sex trafficking,” says Dr Andy Przybylski, an experimental psychologist at the Oxford Internet Institute.
In terms of mental health, says Nicholls, the most important thing is not how much time you spend online but whether that time gets in the way of other things. “We did a longitudinal study looking at whether the time online predicted mental health, but we found it wasn’t time online itself but whether it stopped people getting enough sleep or exercise,” she says.
Meanwhile, van Zalk thinks that the attention to social media and mental health itself is overstated. “My big concern isn’t people’s use of social platforms,” she says. “I find young people very savvy and good at distinguishing between what’s potentially harmful and what isn’t.”
Instead, she worries about filter bubbles: the fact that different people will be given different information, according to what the algorithms think will be most suited to them.
“In the past, if you went to a physical library, you’d troll through actual volumes and find lots of fun stuff that you weren’t looking for,” she says. But now, “we’re all using largely Google and, for better or for worse – it’s probably well-meaning – this algorithm learns from you so it’ll show you information relevant to you.
“When you compare [the results of] two people’s Google searches, they’re wildly different because of their previous searches.”
That, she says, “can be harmful, in the sense that you’re not being exposed, as a young person, to wider sources of information. It’s worrisome because of its effect on democracy.”
The “attention economy”
Digital media “is created to maintain continuous use, exploiting vulnerabilities in the human psyche,” van Zalk told The Forum’s online event. Short-term rewards, in the form of likes or other feedback, are linked, she said, to compulsive behaviour.
“When Facebook added the like button, other networks followed suit,” she told attendees, “and I noted that it was having very interesting effects. It validates you; someone likes you and you feel warm inside. Or, conversely, if someone doesn’t, or you have an idea of what a suitable number of likes is - 'if I don’t get 25 likes for this I’m a worthless human being’ – you start seeing a negative effect on wellbeing.”
These short-term rewards are designed to keep you on the site. “In Silicon Valley these things are called key performance indicators, KPIs,” says Przybylski.
“The way these companies are structured, the KPIs are things like monthly active users, on-platform spend, growth metrics; whatever the investors look for.” The companies want to maximise those KPIs, and one way to do that is by making their site hard to leave, with short-term rewards. Van Zalk calls this the “attention economy”.
Digital media is created to maintain continuous use, exploiting vulnerabilities in the human psyche.
There’s another layer: the things you like, the things you share, the things you search for tell companies things about you; things about your demographics, your preferences, your whereabouts. And the data that they glean from that can be used to sell you things, so it is valuable, and can be sold to advertisers.
“The problem is this surveillance capitalism model is too lucrative,” says van Zalk. “It’s too enticing, they all live off the advertising model. When Google started, they swore they wouldn’t go down the advertising route – but they did, under pressure from shareholders.”
This represents a major obstacle to researching and minimising online harms: the companies themselves. It is in their interests to maximise the amount of time users spend on their site, and the amount of personal information they provide.
Equally importantly, they have huge amounts of data that would be an absolute goldmine for researchers, but they are strongly incentivised not to release it. They want to use that data for their own research.
“I think you can forget about your data from Facebook,” says van Zalk. “Facebook have their own research and aren’t giving it away for free.”
Some researchers have, she says, been given bits of data, but it was “hugely inadequate” for any useful research; Facebook keep the good stuff for themselves.
“I do not foresee a future where Amazon or Google or Apple say ‘Okay, you have all this data and see how we’ve tracked all the things.’ I’d like to think that there’d be some collaboration, and the door is open, but I can’t see it.”
The huge amounts of money flowing into the big tech companies also stifle research, says Przybylski, simply because the tech companies are able to offer promising young researchers many times the salary they could get in academia, and access to all the data they could want.
“The people who run these companies hire, and better pay, our best and brightest,” he says. “Access to the data, the ability to innovate; it’s all tempting to talented young people. It’s literally the plot of Jurassic Park.”
Out of science and into politics
“Online harms” is a necessarily huge and nebulous concept. Unwanted sexual advances online, or grooming, or radicalisation, are entirely different things from Instagram causing body image problems, or Twitter reducing attention spans, if those things are real.
Lumping them together is “a political decision”, says Pzrybylski. “The online world is in many ways as diverse as the offline world. But there’s been a decade of dissatisfaction with the self-regulation of these titanic companies.”
The UK government’s plan to consider all of them as one thing, “online harms”, in their consultation and white paper – and, eventually, by creating an online regulator – is an attempt to create political momentum, he says.
“We’re bundling our fears and dissatisfactions with these companies, and the positions they occupy, into the political will to do something that isn’t reactive. On that basis it’s a promising political exercise, it has potential.”
There’s been a decade of dissatisfaction with the self-regulation of these titanic companies.
This approach has great strengths, he says, but also risks – lumping things together makes them seem the same, so the less well-evidenced areas of potential harm, such as the impacts of social media on wellbeing, could be given the same weight as the better-evidenced ones.
“You get onto claims like pornography harms young people, or violent video games cause violence,” he says. “You can talk about these negative effects without having to have a defensible and falsifiable frame. I’m very sympathetic to it, but it’s open to critiques around free speech, around online censorship.”
And, importantly, science can be in danger of being used as a smokescreen.
“If you decide that photoshopped women on magazine covers are perpetuating an unrealistic ideal,” he says, “you don’t need to do a study. You can just decide: ‘no photoshopped magazine covers’. But the problem comes when you cloak it in evidence.”
But by using words like “harm”, and the language of public health and scientific debate, he says, policymakers can try to hide political decisions behind science.
“You’re making a scientific claim, and you have to be forthright about it. It’s an object for public debate; you and I might disagree, but don’t point at a scientific journal and wait for an academic to settle it for you. Science and data is key, but it’s not a prop, either to sell a pop-psych book or gain a ministerial post.”
What next?
Speaking at The Forum’s online panel discussion, Ali Shah, head of technology policy at the Information Commissioner’s Office (ICO), noted that the internet was never designed for children and young people.
“In 2018 we were tasked by the government to create a code of conduct for age-appropriate design online. What the code tries to address is that the internet wasn’t designed with children in mind, but children are significant users.”
It’s going to be difficult, because the evidence is patchy and research is difficult. “But just because it’s complex and fast-moving doesn’t mean we shouldn’t act,” said Shah.
We need to move past the ‘it’s good’ and ‘it’s bad’ camps, and find ways to protect children.
Working out what is and isn’t safe for young people online is going to be key, and then designing platforms that they can use safely. Van Zalk, with the ICO and Imperial’s Dyson School of Design Engineering, has been working on web and app designs that are suitable for use by young people.
It’s also important to be careful, says Przybylski. “The UK has tried this before with other things they think cause harm, and I am of the opinion that they have failed spectacularly.
“I’d point at gambling: the regulation and research around it are in many ways more simple than online harms, and the harms are more obvious and concrete, but we’re no closer to meaningfully minimising those harms.”
So doing good research is absolutely vital, but so is being honest about what your interventions are, and what a good outcome is.
“If you were running a needle exchange,” Przybylski says, “you’d quantify the harm that’s happening, and then you’d figure out your goal, what you want to bring it down to. How many needles are turned in, how many people turn up in hospital with hepatitis.”
Something similar needs to happen with online harms, he says, and often that doesn’t happen. “With pornography, we had this giant debate about age verification. There was an assumption of harm, but nobody told me the criteria for success: imagine 45% of children had seen it, what percentage do we accept, as a society? How much are we willing to spend to reduce it to 35%?” If we’re using the language of public health, in “harms”, we need to think like health economists.
The most important thing, says van Zalk, is acknowledging that it is a complex, nuanced situation.
“People are trying to tell a coherent narrative that fits the common view,” she says. “And the common view from parents is a fear and a lack of knowledge.”
A lot of the fear, she says, is unfounded – on the Jean Twenge arguments about social media and mental health, “reanalysis of the data shows that the links to depression and anxiety are just not there”.
But there are real concerns, around the attention economy and surveillance capitalism, that we need to address. “We need to move past the ‘it’s good’ and ‘it’s bad’ camps,” she says, “and find ways to protect children.”
The Forum is Imperial’s policy engagement programme. It connects Imperial researchers with policy makers to discover new thinking on global challenges. Our features provide a shop window into the world leading research taking place at Imperial and provide insight into how it can inform and contribute to public policy debates.
Built with Shorthand