Anses Warns Social Networks Harm Teenagers’ Health, Citing Algorithms and Suicide Content

France’s health agency Anses warned that social networks seriously harm teenagers’ health, saying the negative effects are numerous and well documented in an opinion published on Tuesday after five years of work by a multidisciplinary panel of experts.

“If they are not the sole cause of the deterioration of adolescents' mental health, the negative effects of social networks, recently banned for those under 16 in Australia, are numerous and well documented, says Anses in an opinion published on Tuesday, the result of five years of work by a multidisciplinary panel of experts”

ID

Anses said the platforms’ attention-capture systems exploit vulnerabilities inherent to adolescence while they lack the emotional and behavioral regulation capacities of adults, and it linked social media use via a smartphone to sleepiness, irritability, sadness and depressive symptoms.

ID

The agency said content personalization algorithms expose some minors to posts encouraging suicide, self-harm or risky behaviors and to cyberviolence, and it stressed that health-damaging content must be banned.

Anses also tied the danger to AI, warning that “loss of critical thinking” and the “generation of stereotypes or dangerous content” heighten risk, while it argued platforms must modify personalization algorithms, persuasive interface techniques, and default settings.

Anses said the platforms must deploy reliable age-verification systems and collection of parental consent to comply with the European Digital Services Act (DSA) and the General Data Protection Regulation (GDPR), with registration from age 13 and parental validation between 13 and 15.

In California, a historic trial opened on Tuesday to decide whether Instagram, Facebook or YouTube knowingly designed their apps to make youth addicted to social networks, with Judge Carolyn Kuhl overseeing legal duels planned over several months.

The hearing began with a lengthy jury-selection process before debates opened on Monday, February 2, and the case is expected to last six to eight weeks before the Los Angeles Superior Court.

Metricool

Matthew Bergman, founder of the Social Media Victims Law Centers, said the first time social networks have to face a jury for harming children is why the civil case is highly anticipated, and he said the outcome could set a major legal precedent for a wave of US lawsuits.

The plaintiff in the K.G.M. case is a 19-year-old Californian who started using YouTube at age 6 and then opened an Instagram account at 11, before Snapchat two years later and TikTok at 14, and her complaint targets ByteDance, Meta and Alphabet.

The tech giant defense strategy described in the coverage is that Meta considers itself protected by Section 230 of the Communications Decency Act, while Bergman said, “We accuse them of designing their platforms to make children addicted”.

A study discussed by Le Monde.fr said that activating the “For You” feed on X in the United States shifts political opinions in a direction favorable to Republicans, while its deactivation does not lead to a “reversal.”

“What if everything you thought you knew about social media algorithms was completely wrong”

Metricool

The article said the personalized feeds on X, Facebook, Instagram, or LinkedIn replace or supplement a reverse-chronological feed and are designed to maximize engagement through clicks, likes, replies, shares, time spent, and purchases.

Le Monde.fr said the results challenge earlier research, including a large-scale study published in 2023 in Science that found political attitudes are not affected by deactivating recommendation algorithms during the 2020 U.S. presidential election.

Separately, TechCrunch described how social platforms are handing over some power by allowing users to personalize their algorithms with AI, including Threads’ “Dear Algo” tool and its later “Your Algo” feature.

TechCrunch said Instagram head Adam Mosseri has said social media ranking models have historically been built with technology that wasn’t transparent to users, but now large language models (LLMs) can make recommendation systems more understandable by showing why content is displayed and letting users explicitly communicate their preferences.