Digital addictions: what the research supports and what was guesswork
By Brenner Cruvinel • 4 minutes read
content about addiction, pornography, gambling, eating disorders and suicide risk.
This topic matters to me: it was because of a friend that I wrote Hoff's first blocker. For that very reason, the first version of this post had to come down.
It was a table of 29 "dopaminergic addictions", from social media to travel apps, each with a cause, age range, harms and treatment, and every one of them ended in "CBT". When I checked the sources, 24 of the 28 news links returned a 404 error, eight of them with the same invented date in the URL. Of the studies cited, only 4 were about the subject of their row. The one on gambling was about endometrial cancer, and the one on dating apps was about a disease in cattle. It was text that looked like research with no research underneath.
I threw the table out and redid it with what holds up.
What "dopaminergic" means
The post's name comes from a popular idea, that these apps "give you dopamine" like a drug. The science is more interesting than that. In 1997, Schultz, Dayan and Montague showed that dopamine neurons do not signal pleasure itself but prediction error: the difference between the expected reward and the one received. That is why unpredictable reward, like the next notification or the next video, hooks so well. "Dopaminergic addiction" is a nickname, and nothing more.
What is a diagnosis and what is not
The two classifications clinicians use partly disagree:
| Behaviour | ICD-11 (WHO) | DSM-5-TR (APA) |
|---|---|---|
| Gambling | 6C50, disorder due to addictive behaviour | Gambling disorder, the manual's only behavioural addiction |
| Video gaming | 6C51, disorder due to addictive behaviour | In Section III, as a condition needing further research |
| Compulsive sexual behaviour | 6C72, but as an impulse control disorder, not an addiction | No diagnosis of its own |
| Social media, shopping, work | No code of their own | No diagnosis of their own |
Sources: the ICD-11 browser, the APA page on internet gaming and the APA fact sheet on addictive disorders. In ICD-11, gaming disorder has three features: impaired control over gaming, increasing priority given to gaming over the rest of life, and continuing to game despite bad consequences, usually for at least 12 months.
That does not mean the rest is harmless. It means there is still no consensus on where a habit ends and an illness begins.
What the studies say
Video gaming
A meta-analysis by Stevens et al., from 2021, pooled 53 studies with 226,247 people in 17 countries and estimated worldwide prevalence at 3.05%. With only the studies using stricter sampling, it fell to 1.96%. The screening tool explained 77% of the variation between studies, meaning much of the difference comes from how it is measured. Men appear 2.5 times more often than women. The article has a 2023 erratum, which I did not read. On treatment, a 2025 meta-analysis with 21 controlled trials and 1,360 participants found a large reduction in symptoms, with CBT and other approaches working similarly, but with high heterogeneity between studies.
Gambling
The 2012 Cochrane review pooled 14 trials with 1,245 people. Right after treatment, CBT reduced symptom severity and money lost, with effects from medium to very large. But only one study followed people 9 to 12 months later, and there the effect was smaller and not significant. It works in the short term; the long term is still little studied.
Social media and free-to-play games
Montag et al., in 2019, analysed popular apps and described six design mechanisms that prolong use, tied to classic effects from psychology and economics such as the mere-exposure effect, the endowment effect, the Zeigarnik effect and social comparison. Their conclusion is honest: it is very hard to measure the effect of each element without data that only the companies have.
Pornography
The systematic review by de Alarcón et al., from 2019, says it is still unknown how to define when use becomes pathological, because of sample bias and the lack of diagnostic instruments. What they highlight is the "triple A": accessibility, affordability and anonymity, which make problematic use easier.
Shopping
In Black's review, from 2007, compulsive buying has a lifetime prevalence of 5.8% in the US population. About 80% of cases studied in clinics are women, a difference the author himself says may be an artefact. There is no standard treatment; group CBT looked promising.
Work
Andreassen's review, from 2014, concludes that workaholism still lacks clarity as a concept and that no well-evaluated intervention exists.
What was cut
The other 23 rows of the table went: streaming, dating apps, news, messaging, forums, email, fitness apps, podcasts, music, reading, virtual reality, short videos, stalking, meditation, languages, finance, friendship, productivity, diet, travel, art, augmented reality and self-help. None had a study supporting what the row said. The treatment table went too, which listed electroconvulsive therapy and acupuncture next to CBT, with no source linking them to digital addictions.
The question that remains, and that Hoff tries to answer from the practical side, is how to create friction between the impulse and the click without pretending we know more than we do.
If you are struggling, look for a mental health professional. In Brazil, CVV answers 24 hours a day on 188, free of charge; elsewhere, look for your local crisis line.