Alessandro 'kLeZ' Accardo personal website

The personal website of an Italian developer. /nick kLeZ, since forever.

Social media considered harmful. Maybe... (1/5): They knew

Posted on • by

On 1 October the Pew Research Center published a survey, carried out in 37 countries and territories, of more than 45,000 people. The headline says that around the world a growing number of people are convinced social media is harming democracy.

I’ve been convinced of it for ages, and not just where democracy is concerned. I think social media, as it’s built today, is an addictive product, sold by companies that know it full well and that have done what they learnt from those who came before them, namely the cigarette makers and the oil companies: deny and stall, and meanwhile rake in loads of money at the expense of ordinary people. I’m saying it straight away so it’s clear which side I’m on. This is the first of five instalments, one a week, in which I try to prove it with facts, as I usually like to do. The maybe in the title comes at the end, so stick with me until then.

§The survey is just a symptom

I’m starting with Pew because it’s the reason I’m writing now, and because it deserves a better reading than the one given by those who picked it up and wrote the headlines.

The median figure still says the opposite of the headline: 55% of respondents think social media has been good for their country’s democracy, 35% that it’s been bad. What’s growing is the second slice of the pie, and it’s growing faster in rich countries. In four years the “bad” answers have gone from 54 to 70% in the Netherlands, from 41 to 61% in Germany, from 51 to 65% in France. In the United States they’ve been stuck at 64% since 2022. In Italy 54% say good and 44% say bad, with one detail: 88% of Italians think social media has made people easier to manipulate with false information, against a median of 79%.

The same people, in the same survey, also say social media has made them better informed (77%). Keep that in mind, because it’s exactly what you’d expect from a product that gives you something while taking something else.

Then there’s a figure Pew reports without comment: the youngest and those who use the most platforms are the most positive. My reading, in my very humble opinion, is that the least critical are the most exposed. If that rings a bell, it’s because that’s usually how it goes with things that get you hooked.

A survey, though, measures an opinion. People starting to smell smoke doesn’t prove the house is on fire. For that you need documents, and the documents exist, oh boy do they exist. But first I have to tell you where the script comes from, and it’s much older than Facebook.

§Tobacco wrote the script

In 1959 the director of the California Department of Public Health wrote that the evidence had become conclusive: cigarette smoking is a major cause of lung cancer. In 1961 a World Health Organization study group reached the same conclusion. In 1963 the same Californian department put numbers into an official report: the risk of dying of lung cancer was 14 times higher for smokers, 26 times for those smoking more than two packs a day. In January 1964 came the report of the US Surgeon General (stretching it a lot, because there’s no real equivalent here, it’s a sort of head of the national health institute, but appointed by the US President for a four-year term).

Ten years after that first “conclusive”, in 1969, someone at Brown & Williamson put the company’s strategy down in black and white in an internal document titled Smoking and Health Proposal:

Doubt is our product since it is the best means of competing with the “body of fact” that exists in the mind of the general public. It is also the means of establishing a controversy.

Read it again, because it’s one of the most honest sentences ever written by a company, and it was written so that nobody outside that room would ever read it.

It worked for decades. While the science was settled, cigarettes sponsored Formula 1, and in Europe they were allowed to until 2005, when the directive banning cross-border sponsorship came into force. In the United States the bill arrived in 1998, with an agreement between the four biggest manufacturers and 46 states: at least 206 billion dollars over twenty-five years. And in 2006 a federal judge, Gladys Kessler, ruled, in a civil case brought under the organised crime law, that the companies had deceived the public about the health effects and addictiveness, and that they had marketed their lethal product “with zeal, with deception”.

The internal papers, including that sentence, only came out in May 1994, when about 4,000 pages of Brown & Williamson documents arrived in the post for Stanton Glantz, a professor at the University of California. From that parcel to the agreement with the states: four years. To the ruling: twelve. Write those numbers down.

§Then oil performed it

Between 1977 and 2003 Exxon’s scientists produced, for internal use, a series of global warming projections. In 2023 three researchers set them one by one alongside what actually happened and published the result in Science: the internal projections forecast warming of 0.20 degrees per decade, give or take 0.04. They were right.

Meanwhile, in public, the company said the science was uncertain.

The mechanism is the same, and it has four moves. Do serious research in-house, because you need to know. Find out your product does damage. Keep the research in a drawer. Outside, sell doubt. The beauty of it is you don’t even have to lie about everything: you just have to say “more studies are needed”, which is always true.

§They knew too

On 8 November 2017, in Philadelphia, Sean Parker, Facebook’s first president, let his guard down in front of an audience and an Axios journalist. He said the thinking behind how they built the product was: how do we consume as much of your time and attention as possible? He talked about a mechanism that rewards you with other people’s approval and exploits a vulnerability in human psychology. And he added the part that matters:

The inventors, creators […] understood this consciously. And we did it anyway.

That was a spoken confession. The paperwork arrived in September 2021, when the Wall Street Journal published the internal documents that Frances Haugen, a former employee, had taken out of the company. They were the research Facebook had done on Instagram and teenagers. A 2019 slide said:

We make body image issues worse for one in three teen girls.

Another reported that teenagers blamed Instagram for increases in anxiety and depression, and that the reaction was “unprompted and consistent across all groups”. Yet another that, among teens who had reported suicidal thoughts, 13% in the UK and 6% in the US traced them back to Instagram. The material, the paper writes, had been reviewed by top executives and cited in a presentation to Mark Zuckerberg in 2020.

Now the part that people writing pieces like this usually skip, and that I’m putting in, because if I skip it it’ll be thrown back in my face. Meta responded by publishing the slides with its own annotations, and the annotations say true things. That “one in three” refers to girls who already had body image issues, not all teenage girls. Fewer than 150 girls answered the body questions. On eleven of the twelve topics on that slide, the company says, girls who were struggling more often said Instagram helped them, and body image was the only exception. The most common answer was “no effect”. And that kind of research can’t prove causation.

All fair. These are small studies, done to steer a product, and they wouldn’t survive peer review.

One thing remains, and it’s the same as with tobacco. In March 2021, while those slides were sitting in the company’s drawers, Zuckerberg told a congressional committee:

The research that we’ve seen is that using social apps to connect with other people can have positive mental-health benefits.

It’s a half-truth, and it’s the convenient half. A year earlier his researchers had also shown him the other half: for girls already struggling with their bodies Instagram made things worse, and teens themselves blamed it for anxiety and depression. Only the good part made it to Congress. You can argue your internal studies are weak, or you can cite them in your favour before Parliament. Not both at once. Strictly speaking he didn’t tell a lie, because the sentence was technically true, but it wasn’t the whole truth. And nobody asked him for it: at that hearing the witnesses weren’t sworn in, so much so that a congressman, Jeff Duncan, protested on the floor that “They should be sworn in”, and they carried on regardless. It’s called omertà: knowing, and keeping quiet about the part that hurts you, in front of a Parliament. And Congress, which answered Duncan’s protests by changing the subject, was complicit and compliant.

There’s one episode in those same documents that, to my mind, is worth more than all the percentages. It was called Project Daisy: hiding the like count under photos, to ease the pressure on kids. The company tested it, and its researchers reported to Zuckerberg that they hadn’t seen any improvement in well-being. They launched it anyway, in May 2021, as an option. The reason is written in the documents: it would be welcomed by the press and by parents as proof that Instagram cares about its users.

A fix you know is useless, put in the shop window because it looks good. If you’re after an operational definition of selling doubt, there it is.

On 5 October 2021 Haugen testified before the Senate:

The company’s leadership knows how to make Facebook and Instagram safer, but won’t make the necessary changes because they have put their astronomical profits before people.

Senator Richard Blumenthal, commenting on those documents, said Facebook seemed to have taken a page from the Big Tobacco playbook. So I’m not the first to make the comparison.

And on 31 January 2024, at another Senate hearing, with the parents of children who had been exploited, bullied or pushed to self-harm in the room, holding up photos of their kids, Zuckerberg stood up, turned to face them and said he was sorry for everything they had been through. Then he sat back down and went back to talking about the company’s investments in safety.

Had I been one of those parents, I don’t know if I’d have been quite so composed. Maybe in their shoes I’d have gone for his jugular like in Dawn of the Dead, and today Hegseth would be announcing the livestream of my death, hoping they hit the vein and don’t land me in hospital instead.

§How the little machine works

So far I’ve said they knew. What’s left is to say what they knew, that is, how an app keeps you hooked.

The principle is seventy years old and comes from psychology labs: it’s called variable-ratio reinforcement. If a lever gives a reward exactly every ten presses, the animal takes a break after each reward. If it gives it at random, sometimes on the second press and sometimes on the fiftieth, the animal presses at a high, steady rate, and keeps going for a long time even after the reward stops coming. It’s why slot machines work, and the anthropologist Natasha Dow Schüll wrote an entire book on it, Addiction by Design, after spending years studying Las Vegas casinos and the people who stay glued to them.

Now look at your phone. You pull down to refresh and you don’t know what’s coming. You scroll and there’s no bottom. The notification could be your mum, a like, or nothing to do with you. Every gesture is a lever, and the reward comes at random.

One caveat, because there’s a lot of hot air going around on this topic. The story about likes releasing dopamine “like cocaine” I’m leaving out: it’s an oversimplification that neuroscientists dispute. The behaviour is measurable, and that’s enough for me.

The US Surgeon General laid it all out in 2023, the very same office that signed the smoking report in 1964. Its advisory on social media and young people lists the features designed to maximise time spent: notifications, autoplay, infinite scroll, likes, recommendation algorithms. And it reports some numbers:

  • up to 95% of 13- to 17-year-olds use social media, more than a third “almost constantly”;
  • those who use it for more than three hours a day face double the risk of symptoms of depression and anxiety (a study of 6,595 adolescents aged 12 to 15, which accounted for how they were doing beforehand);
  • nearly a third of social media use, 31%, may be attributable to self-control problems magnified by habit (estimated by a study by three economists, titled without beating about the bush Digital Addiction).

The same document uses the language of addiction, and does so cautiously: some researchers believe exposure to social media may activate circuits comparable to those of addiction, and small studies have found brain changes in problematic users similar to those of people addicted to substances or gambling. I’m keeping the caveats, because they’re theirs. The fact remains that a federal health authority wrote those words.

The Surgeon General’s overall conclusion is a sentence nobody would accept for a medicine or a child’s car seat:

At this time, we do not yet have enough evidence to determine if social media is sufficiently safe for children and adolescents.

And meanwhile up to 95% of kids use it.

In Europe the World Health Organization measures problematic social media use among adolescents every four years, the kind with symptoms of addiction: loss of control, withdrawal, the rest of life taking a back seat. It was 7% in 2018. In 2022 it was 11%. In Italy, according to the Istituto superiore di sanità, we’re at 13.5%.

§It should be in the manual

The manual is the DSM, the Diagnostic and Statistical Manual of Mental Disorders, the book psychiatrists across half the world use to decide what is a disorder and what isn’t.

Social media addiction isn’t in it. But there is a precedent, and it’s not a small one. In 2013 the fifth edition moved pathological gambling into the addictions chapter, alongside alcohol and substances: the first recognised behavioural addiction, meaning one without a molecule involved. In the same edition internet gaming disorder made it onto the list of conditions requiring further study. The World Health Organization went further: in its classification of diseases, in force since January 2022, gaming disorder is a diagnosis in its own right.

So the idea that a behaviour can cause genuine addiction is already written in the manuals. That a digital product can do it, likewise. And the first one to be recognised was precisely the one social media copied its mechanism from: gambling.

In my view social media addiction should be on that list. It’s an opinion, sure, I’m not a psychiatrist and I’m not the one writing the manual (obviously, the clever folk would say “ça va sans dire”). But all the pieces are there: the mechanism, which is gambling’s, the symptoms, which the WHO has been measuring for years with a dedicated scale, and an industry that studied the phenomenon in-house and downplayed it outside. The last time we saw these three things together, we were talking about nicotine.

I know there are people who disagree, and they’re serious people. There are researchers who find small effects across large samples, and who urge us not to panic. One of them, Andrew Przybylski of Oxford, said something I find perfect:

People talk about Instagram like it’s a drug. But we can’t study the active ingredient.

He’s right. And we can’t study it because the data belongs to the manufacturer, who opens it up to whomever it likes, whenever it likes and however it likes. In the sixties, at least, you could buy a cigarette and take it to a lab.

I’ll get to the sceptics in my own time in the next instalment, along with the courts, the laws and the lobbyists. Because in the meantime somebody has made it in front of a jury, and it didn’t go the way you’d expect.

With tobacco, twelve years passed from the parcel of documents to the ruling. The Facebook Files are from 2021. The internet speeds everything up, and I hope it speeds this up too.

§Sources