The Experiment That Rewired How We Understand Social Media Peer Influence on Teenagers
In a study that has since become one of the most cited pieces of evidence in debates about social media's impact on young minds, a team at the University of California, Los Angeles placed teenagers inside a functional MRI (fMRI) scanner and showed them photographs — some mundane, some risky — each stamped with a like count secretly assigned by the researchers themselves. The results, reported by Silicon Canals, were striking: teenagers consistently preferred to like photos that already appeared popular — regardless of whether the photo showed a basket of fruit or something considerably more risky. Social media peer influence on teenagers, in other words, operates almost independently of content.
The UCLA team, led by developmental neuroscientist Dr. Lauren Sherman, published their findings in the journal Psychological Science. Their conclusion was clinically precise: the number of likes a photo displayed was a far stronger predictor of a teenager's decision to like that photo than the actual content of the image. The brain scans reinforced this, showing heightened activation in regions associated with reward processing and social cognition — including the nucleus accumbens — when teenagers viewed highly liked photos. For privacy professionals, platform designers, and policymakers, those neural signatures carry significant implications that stretch well beyond adolescent psychology.
What the Brain Scans Actually Showed — And Why Platforms Already Knew
The fMRI data revealed something that social media platforms arguably understood through behavioural analytics long before neuroscientists confirmed it with brain imaging: social proof is a more powerful engagement driver than content quality. When a teenager saw a photo with a high like count, the brain's reward circuitry lit up in much the same way it responds to monetary rewards or social approval in face-to-face interactions. What made this experiment methodologically rigorous — and ethically provocative — was that the like counts were entirely fabricated by the research team. Content was held constant; only the social signal changed.
This design stripped away one of the most common defences offered by platform companies: that engagement simply reflects genuine user preference. Here, the preference was manufactured by a number on a screen, not by any quality intrinsic to the image itself. For IT decision-makers and data ethics professionals, this finding cuts to the heart of a persistent question in algorithmic product design: when a platform surfaces content ranked by engagement metrics, is it reflecting user preference, or actively constructing it?
"The like count functions as a social norm signal. Teenagers aren't just evaluating images — they're evaluating what their peers appear to have already endorsed."
— Dr. Lauren Sherman, UCLA Department of PsychologyResearch published in Nature Human Behaviour has similarly documented how social influence cascades shape online behaviour across age groups, not just among teenagers. But adolescents appear particularly susceptible because the prefrontal cortex — the brain region responsible for impulse control and independent judgement — is not fully developed until a person's mid-twenties. In practical terms, this means the very users most targeted by engagement-optimised platforms are also the least neurologically equipped to resist them.
Why This Matters for GDPR Compliance and Age-Appropriate Platform Design
For privacy professionals and compliance teams operating under the General Data Protection Regulation, the UCLA findings provide scientific grounding for legislative instincts that have been building across European institutions for years. GDPR's Article 8 already mandates parental consent for processing children's personal data below the age of 16 (or lower, depending on member state implementation). But the regulation does not directly address the mechanics of engagement design — how platforms use aggregated behavioural data to engineer the very feedback loops the UCLA study documented.
The UK's Age Appropriate Design Code (also known as the Children's Code), enforced by the Information Commissioner's Office, goes further by requiring platforms to default to privacy-friendly settings for users likely to be under 18, and to assess the risks of nudge techniques — design patterns that push users toward engagement-maximising behaviours. The UCLA study provides exactly the kind of empirical evidence that regulators cite when classifying like counts and social proof signals as potential nudge mechanisms. According to the ICO's Children's Code guidance, platforms must not use techniques that are detrimental to children's wellbeing — a standard that, in light of this research, creates real compliance exposure for any service displaying aggregated social approval metrics to minors.
The European Union's Digital Services Act (DSA), which entered full enforcement for large platforms, adds another regulatory layer. Under the DSA, very large online platforms are required to conduct systemic risk assessments specifically covering risks to minors, and to implement mitigation measures for recommendation systems that may have negative effects on users' physical and mental health. The UCLA data — showing that social proof signals override content judgement in teenage brains — is precisely the kind of evidence a DSA risk assessment should engage with. Platforms that have not factored this research into their mandatory risk frameworks may find themselves poorly positioned in regulatory scrutiny.
How Algorithmic Recommendation Systems Amplify What the UCLA Study Found
The UCLA experiment simulated a static like count. Real platform algorithms are considerably more dynamic — and more optimised. Modern recommendation systems on major social platforms do not merely display engagement metrics; they actively surface content that is already performing well by those metrics, creating a compounding feedback loop. A post that accumulates early likes is shown to more users; those users, primed by the high count, are more likely to like it; which drives further distribution. The UCLA study identified the psychological mechanism at the individual level. Algorithmic amplification industrialises that mechanism at population scale.
Research from the MIT Media Lab, referenced in coverage by Wired, demonstrated that false information travels faster and farther on social networks than accurate information — in part because novelty and emotional charge drive engagement, and engagement metrics then signal further distribution. The intersection of that finding with the UCLA neuroscience is uncomfortable: platforms are optimising for the very metric that the UCLA study showed bypasses critical content evaluation in teenage brains.
For developers building consumer-facing applications, this creates both an ethical and a product design question. Features like visible like counts, follower numbers, and trending labels are not neutral UI elements — they are, according to the neuroimaging evidence, active inputs into users' decision-making processes. The question of whether to surface those signals to younger users, and in what form, is no longer a purely commercial choice. It is increasingly a regulatory compliance question, particularly in European markets where digital wellbeing obligations are being written into law.
Platform Responses to Like Count Research — And Why Self-Regulation Has Limits
Instagram itself, acknowledging the psychological weight of like counts, began testing hidden likes in several markets. The experiment — where like totals were visible only to the post's author, not to viewers — was framed as a wellbeing initiative. But the rollout was inconsistent, eventually becoming an opt-in feature rather than a default setting. Critics, including digital rights organisations such as the Electronic Frontier Foundation, argued that making harm-mitigation features optional rather than default reversed the protective logic: users who most need protection are often the least likely to proactively enable it.
This tension — between platform commercial incentives and user wellbeing — sits at the core of the ongoing regulatory debate in Europe. The DSA's risk assessment requirements are deliberately structured to prevent platforms from treating known harms as acceptable externalities. Regulators have made clear that self-regulatory gestures, such as optional like-hiding, do not satisfy the obligation to systematically mitigate risks. The UCLA study's findings give that regulatory logic a firm empirical foundation: when peer validation signals are visible, they demonstrably alter behaviour in ways that bypass content evaluation. That is not a hypothetical risk — it is a documented neural mechanism.
| Regulatory Framework | Jurisdiction | Key Obligation Related to Teen Engagement | Enforcement Body |
|---|---|---|---|
| GDPR Article 8 | EU / EEA | Parental consent for data processing of under-16s | National Data Protection Authorities |
| Age Appropriate Design Code | United Kingdom | Default privacy settings for under-18s; ban on detrimental nudges | Information Commissioner's Office (ICO) |
| Digital Services Act | EU | Systemic risk assessment for minors; algorithm transparency | European Commission / DSA Coordinators |
| AI Act (pending full application) | EU | Restrictions on AI systems that manipulate behaviour subliminally | National Market Surveillance Authorities |
