For many years, the debate about social media and teen mental health has centred on one key issue: correlation versus causation. While large-scale studies showed that increased social media use was linked with higher levels of anxiety, depression, and self-harm among teenagers, critics argued that this did not prove social media was the cause.
That argument is now increasingly difficult to maintain.
Psychologist Jonathan Haidt argues that the evidence has moved beyond correlation and now demonstrates causation. He has stated that there are now “tons and tons of evidence of causation” showing that social media platforms are directly contributing to harm.
New Evidence Supporting Causation
The updated evidence is presented in a paper titled Mountains of Evidence, which brings together findings from multiple independent sources. Rather than relying on a single type of study, the research draws on several different lines of evidence that point in the same direction.
Meta’s Internal Research
One major source of evidence comes from internal studies conducted by Meta. These studies became public through whistleblowers and legal cases brought by US state attorneys general.
Some of these internal studies were true experiments. They showed that when users reduced or stopped using Facebook or Instagram, measures of depression, anxiety, and loneliness decreased. This indicates that the platforms themselves were aware of harmful effects, even while publicly downplaying them.
Experimental Studies With Random Assignment
Independent experimental studies provide further support. In these studies, teenagers were randomly assigned to reduce or stop social media use for a short period, often a week or more.
Across multiple experiments, researchers found that reducing social media use led to measurable improvements in mental health, including lower levels of anxiety and depression. Random assignment is important because it helps rule out alternative explanations, strengthening the case for causation.
A Broader Range of Harms
Earlier discussions often focused narrowly on social comparison, such as comparing appearance or popularity. The newer research highlights a wider range of harms, including:
- Sextortion, described as one of the clearest and most direct harms
- Sexual harassment, with internal platform research suggesting that around 15% of teens experience it in a given week
- Cyberbullying
- Exposure to violence
- Exposure to extreme or hardcore pornographic material
These harms do not rely on subjective interpretation and are linked directly to platform features and online behaviour.
Separating Two Different Questions
Haidt argues that confusion in the debate comes from failing to separate two distinct questions.
The Historical Question
The first question asks whether social media caused the large rise in teen mental health problems observed after around 2012. This question is difficult to answer with absolute certainty because it is not possible to run controlled experiments on entire generations over time.
Haidt acknowledges that this question cannot be answered with 100% confidence.
The Product Safety Question
The second question asks whether social media, as a consumer product today, is harming children and adolescents. On this question, Haidt states that he is 99.9% confident that it is.
This confidence is based on seven different lines of evidence, including correlational studies, longitudinal research, experiments, and internal company data. Together, these form a consistent pattern showing that social media use causes harm for a significant proportion of young users.
Haidt argues that this is the question that matters most for parents, schools, and policymakers, as it directly concerns child safety in the present.
References
Haidt, J., & Rausch, Z. (2024). Mountains of evidence: Why the case against social media has only grown stronger. After Babel. https://www.afterbabel.com/p/mountains-of-evidence
The New York Times. (2026, January 16). Jonathan Haidt on the new evidence linking social media and teen mental health [Audio podcast episode]. https://www.nytimes.com/2026/01/16/podcasts/jonathan-haidt-new-evidence.html
