Psychologists Discover Downside of Social Media Algorithms

In the study, 60 young adults with an average age of 20 were assessed via validated anxiety and depression scales. They then watched two sets of videos: one of personally recommended videos from their own Instagram or TikTok accounts, which were further categorized for content types; the other of generalized trending videos. Brain activity was recorded using EEG to measure frontal alpha asymmetry - a well-established measure of emotional processing and motivation.

"In frontal alpha asymmetry, dominance on the left represents more positive emotional processing," Tang said. "Dominant right-sided activity is usually related to more negative processing and withdrawal-related behaviors.

"Instead of after-the-fact reflections, we looked at real-time emotion processing and documented reactions to watching personalized recommendations. In that regard, this study is novel. Prior research on this has been based on survey questions and self-reports. We are reporting what the study participants are looking at and how they are affected by what their feeds serve them."

The study accounted for total screen time as a confound, or outside factor, which is used in the majority of research on this subject. First author Carole Leung, a doctoral student in cognition and neuroscience, explained that screen time might have weak relations with mental health because that time is not uniformly negative.

The study results confirmed that participants with more depressive symptoms are provided recommendations that lean more depressive, and their brain activity - as determined by frontal alpha asymmetry - is characteristic of depression. This is distinct from their brain activity while watching videos that are trending generally.

"What we're finding is that if you're already feeling depressed, then the algorithms reflect that. Continued engagement reinforces this negative feedback loop," Warren said. "Participants bring in their own phones; we are not selecting videos for them. It's a new take on understanding how social media consumption affects people."

The most common category of videos recommended to young adults concerns social relationships, either romantic or friendships. Tang said researchers are finding that the feelings participants displayed could impact the apps' recommendations.

"If you're more depressed, you're more likely to be exposed to a video showing an argument or conflict," she said. "Positive videos would be about friends supporting each other, for instance. There are videos we classify as neutral as well - those lacking emotional connotation."

Viewing fewer positive social relationship videos also was associated with relative right frontal alpha asymmetry, or more negative feelings.

"Depressed individuals have a tendency to focus on negative things, dismissing the positive things they encounter," Warren said. "If you're depressed and you keep on scrolling, you will encounter fewer positive videos."

Researchers hope to highlight ways to develop healthy social media engagement behaviors so that negative mood isn't amplified. The main intervention now is educating people on how to engage with social media more critically and effectively.

"We are teaching teenagers who frequently use TikTok and Instagram how to curate positive content while limiting the negative by following and interacting with the right creators," Tang said. "If they are tired of their feeds, an account can be set back to default, providing brand-new recommendations. From there, you can reteach the app what interests you."

Parents also can be more aware of how algorithms influence mood, but limiting teens' time on social media won't teach them how to respond in a useful or adaptive way.

"Two years ago, the guidance from the American Academy of Pediatrics was to reduce screen time," Tang said. "These days, if you tell your teenager not to go on their phone, it's like you're taking away their life. That's not going to work."

Researchers are now collecting data on a cohort of teens ages 13 to 16 for a potential long-term study to understand the emotional effects of algorithm suggestions over time.

Other UT Dallas-affiliated contributors to the current study include Lucie H. Nguyen BS'25, an incoming medical student at Dell Medical School at UT Austin; former UT Dallas research assistant Christina Vlahakos, now a graduate student at Northwestern University; and Dr. Carlos Busso, now at Carnegie Mellon University.

This research was supported by a 2025 Social Sciences Seed Grant from UT Dallas and a Behavioral Health Research Award from the University of North Texas' Center for Psychosocial Health Disparities Research.