Molly Russell, and the rest of the iceberg
Plus the AI tsunami roundup: illustration shifts towards video
Last week a British coroner delivered a remarkable verdict on the death of 14-year-old Molly Russell, a schoolgirl who took her life in November 2017 after viewing depressing pictures and videos on Instagram and Pinterest. In the coroner’s hearing, representatives from the companies were called as witnesses, and asked whether they thought that some of the content she was shown was appropriate for someone her age. As the New York Times noted,
A child psychologist who was called as an expert witness said the material was so “disturbing” and “distressing” that it caused him to lose sleep for weeks.
The verdict was that
“Molly Rose Russell died from an act of self-harm while suffering from depression and the negative effects of online content,” said the coroner, Andrew Walker. Rather than officially classify her death a suicide, he said the internet “affected her mental health in a negative way and contributed to her death in a more than minimal way.”
It’s believed to be the first time that “online content”, particularly social media, has been cited as a cause of suicide.
There’s been lots written on the inquest, and on Molly and the devastating effects on her family—her father Ian has become a campaigner for controls on platforms—and so I want instead to look at what all this must imply for all the other young users out there.
TL;DR
The original draft of Social Warming did contain a whole chapter about what we might be able to figure out about the effects of social media on children. (It was cut because the book was long enough without it, and space is limited.) Molly Russell’s case was already a cause célèbre, and that was how I opened the chapter.
Logically, though, if Molly Russell could be so influenced by what she saw, then there must be thousands, perhaps millions of other children out there who are also being shown depressing content, algorithmically amplifying their woes and reflecting them back to their users. It might not seem like warming, but it’s the same process: the platforms making bad things worse through the indifferent application of algorithms working to maximise attention, and insufficient moderation.
The question was, where would I find data that might confirm this? It turns out there’s a really interesting group of datasets which go back to 2000, and are carried out every three years, examining children and their attitudes to their home and friends’ life. The first is PISA, the OECD's Programme for International Student Assessment. Officially, “PISA measures 15-year-olds’ ability to use their reading, mathematics and science knowledge and skills to meet real-life challenges.” But it also asks them questions about their home life, and as time has gone on, about how much time they spend on the internet at school and at home, and (separately) about their happiness.
There’s an absolute ton of data in PISA, and you could spend days poring over it. I did. To set sail across the sea of PISA spreadsheets is to discover an ocean where the wind only blows in one direction: more internet use, less time spent offline, and unhappier pupils. Correlation isn’t causation—just because two things rise or fall together, or in strict inversion, doesn’t mean that one causes the other—but there is no other pair of indices across the years and the spreadsheets that tells the same story so relentlessly. As I roamed across the rows and columns harvested from thousands of children in hundreds of schools around the world, the story was always the same—so much so that if one of my calculations showed a fall in internet use over a time period, my first suspicion (always confirmed) was an error in the calculation, rather than an anomaly in the data.
Here’s just one graph to consider. The more time children spent online, in all these different countries, the less happy they reported themselves to be.
The other is the US Happiness Index. This has been running since 1973, and asks Americans of all ages a number of questions, but the key on is: “Taken all together, how would you say things are these days—would you say that you are very happy, pretty happy or not too happy?”
The overall picture shows that since 1973, when the survey started, Americans of all ages have become generally gloomier: since 2003, there has been a sharp drop, with self-reported levels never rising above the lowest figures recorded in the previous 30 years. Among teenagers, decades of gradually growing optimism were followed by a sudden drop after 2011. Thirteen-year-olds who spent 10 or more hours per week on social media were significantly more likely to say they were unhappy, while those who spent more time than the average physically with their friends were significantly less likely to say that. Seen alone, the data could also be interpreted as showing that unhappy children spent more time on social media. However separate research showed that the feelings only flowed in one direction: that being unhappy didn’t lead to more Facebook use, for instance. The picture that emerged was that more social media use lowered happiness levels.
Girls, more than boys, seem to shoulder most of the unhappiness at being left out. Depressive symptoms among girls rose by 50% between 2012 and 2015; for boys, the figure was 21%.
There is one part of the US’s Happiness index that researchers haven’t remarked on, but which points to the particular difficulties of adolescence. Because the Index questions the same age groups from across the country over four years—in 8th grade (aged 13-14), 10th grade (15-16) and 12th grade (17-18), it’s possible to track happiness levels within a cohort as they age: if you look at the self-reported happiness levels of 12th-grade respondents from two and four years before, you can build up a picture of how their happiness has fluctuated over time.
The first point that emerges is that in every year, teenagers at every age say they’re less happy than adults. The difference is significant, and though the gap has been smaller since the Great Recession beginning in 2007, it’s still there. Here’s the graph:
The second point, which may seem surprising, is that until 2011 children would always end their teenage years happier than they began them. But since 2011 their final reported happiness score has been lower when they finish than when they begin secondary education. The graph measures the difference between reported happiness from the group at year X, and then at year X+4 (so it’s the same cohort, even if not the same people; as the sample is statistically valid, it should be OK):
The third point is that since 2010, self-reported happiness drops in the middle of teenagerhood—the 15-to-16 ages. This has become marked, and substantial. (The graph that shows this is really messy, so I haven’t included it. But believe me, it’s there.)
To some, these sorts of graphs and sets of data show that things are bad, and it’s all the fault of smartphones and social media. Others really aren’t convinced: the standard objection, that correlation isn’t causation, is hard to refute. And it’s quite easy to think that there’s a different reason why “happiness” might have fallen at the same time as the smartphone and social media appeared: it intersects with the GFC (Great/Giant Financial/Fiscal Crash/Crisis). The calamitous effect of that on jobs and earnings might just have had an effect on adult happiness, which might get transmitted to their children. No need for smartphones.
Thus you get warring headlines like these:
Of course it’s confusing. The researchers themselves can’t agree. And yet I couldn’t find any place where more internet and more smartphone use was followed by more happiness. Molly Russell seems like the tip of a giant, unhappy iceberg. Let’s hope she remains an isolated case.
Glimpses of the AI tsunami
(Of the what? Read here.)
• Tell Stable Diffusion the start and endpoints, and this will create an animation between them. Wow.
• Google has announced Imagen Video, a text-to-video (shall we call them T2V?) system that outputs 1280x768 video at 24fps. Currently in research, but as this is only six months after OpenAI showed offer Dalle-2 (text-to-image, T2I). Woooow. More detail on how it works at Ars Technica, which also links to the research paper.
• Bruce Willis (well, his PR people) denied that he has licensed his face to a company called DeepCake, which makes deepfake adverts and videos. Sure, his face got plastered onto some other actor’s for a Russian phone company ad in 2021, but that (the suggestion is) was a one-off. Willis has stopped doing in-person acting after developing aphasia, which made the original story seem credible.
• Copyright implications of AI illustration systems? There probably aren’t any, because the training falls under fair use/fair dealing (US phrase/UK phrase), rather as happened with Google’s scanning of books for the Google Books project. That’s the analysis of Andres Guadamus, at least. He’s not a copyright lawyer, but makes plenty of salient points. The fact that I completely agree with him on everything is neither here not there, m’lud.
• GhostlyStock. A stock image generator using AI. Hit and miss.
• You can buy Social Warming in paperback, hardback or ebook via One World Publications, or order it through your friendly local bookstore. Or listen to me read it on Audible.
You could also sign up for The Overspill, a daily list of links with short extracts and brief commentary on things I find interesting in tech, science, medicine, politics and any other topic that takes my fancy.