Why the unit of randomization could change the answer
Randomization decides who gets the intervention and who does not. In an individually randomized trial, each student is assigned separately, so classmates may end up in different conditions. In a peer-group cluster-randomized trial, whole groups such as teams or classes are assigned together, so everyone in a group shares the same condition.
That difference matters because students in the same peer group talk to each other. A pilot trial of AI-generated personalized sleep feedback in high school female soccer players randomized individuals within the same team and explicitly flagged a risk from doing so, noting that in multi-team studies cluster randomization at the team level should be considered to reduce that risk [2]. The concern is about contamination between conditions, not about sleep feedback itself, so it is a design lesson rather than a measured effect size.
The social media reduction study does not settle the comparison
One study looks at screen use, physical activity, and sleep among 4,421 Danish adolescents and their link to mental health, and it reports on a social media reduction intervention with sleep and mental health outcomes [1]. It also describes how these behaviors cluster differently by gender, with girls spending time on social media and boys on gaming [1].
What it does not provide is any detail on whether participants were randomized individually or by peer group [1]. Without that, the study cannot be placed on either side of the comparison, and it cannot tell us whether the two designs produce different effect sizes for sleep or mental health.
What can honestly be concluded about effect sizes
No effect-size comparison between the two designs appears in these studies. The sleep-feedback pilot used individual randomization and measured sleep and mood, but it did not test a social media reduction and did not use peer-group cluster randomization [2]. The Danish adolescent study reports a social media reduction intervention but leaves the randomization method unspecified [1].
So the honest answer is that the question remains open in this evidence. The design concern raised by the pilot trial suggests a plausible reason the two approaches might diverge, since within-team individual randomization risks peers in different conditions influencing one another [2], but that is a rationale for caution, not a demonstrated difference in effect size.
Sources used in this answer
… use, physical activity, and sleep among adolescents: A latent class analysis of 4,421 Danish adolescents' movement patterns and the correlation with mental health
Examines screen use, physical activity, and sleep clustering among 4,421 Danish adolescents and their correlation with mental health, reporting on a social media reduction intervention with sleep and mental health outcomes but giving no detail on individual versus peer-group cluster randomization.
Feasibility and Preliminary Effects of AI-Generated Personalized Sleep Feedback in High School Female Soccer Players: Pilot Randomized Controlled Trial
A pilot randomized controlled trial of AI-generated personalized sleep feedback in high school female soccer players used individual randomization within the same team, measured sleep and mood, and noted that team-level cluster randomization should be considered in multi-team studies to reduce risk from within-team randomization.