Social media algorithms have become highly effective at matching users’ viewing habits with personalised content, but new research suggests those recommendations may also be linked to mental health. Psychologists at The University of Texas at Dallas examined how algorithm-selected social media content relates to emotional processing in the brain and symptoms of depression.

The proof-of-concept study, published in Computers in Human Behavior, was led by Dr. Alva Tang, assistant professor of psychology in the School of Behavioral and Brain Sciences. Researchers recruited 60 young adults with an average age of 20 and assessed them using validated measures of anxiety and depression before examining their responses to social media videos.

Participants watched two sets of videos. One consisted of personalised recommendations drawn from their own Instagram or TikTok accounts, while the other featured general trending videos. Researchers recorded participants’ brain activity using electroencephalography (EEG), focusing on frontal alpha asymmetry, a measure associated with emotional processing and motivation. Greater left-sided activity is generally associated with more positive emotional processing, while greater right-sided activity is linked to negative processing and withdrawal-related behaviour.

Unlike previous studies that have relied largely on surveys and self-reported social media habits, the researchers were able to examine emotional processing as participants viewed recommended content in real time. The study also accounted for overall screen time, allowing researchers to look beyond how long participants used social media and focus instead on what their algorithms showed them.

The results indicated that participants with more depressive symptoms tended to receive recommendations containing more depressive content. Their brain activity while viewing personalised recommendations also showed patterns associated with more negative emotional processing. Importantly, these patterns differed from participants’ responses when they watched generally trending videos, suggesting that personalised feeds may play a distinctive role in users’ emotional experiences.

Researchers described the findings as evidence of a potential feedback loop. A user who is already experiencing depressive symptoms may interact with certain types of content, leading the algorithm to recommend more material with similar emotional characteristics. Continued engagement with that content could then reinforce negative emotional responses. The researchers stressed that participants used their own phones and feeds, meaning the research team did not select the personalised videos they viewed.

Social relationships, including friendships and romantic relationships, were the most common subjects among videos recommended to the young adults. Participants with more depressive symptoms were more likely to encounter negative relationship content, such as arguments or conflicts, while viewing fewer positive relationship videos was associated with brain activity reflecting more negative emotional processing. The researchers noted that people experiencing depression may also be more inclined to focus on negative information while overlooking positive material.

The researchers hope the findings will encourage healthier and more deliberate social media habits rather than simply focusing on reducing screen time. They suggest users can influence their recommendations by following and interacting with positive content and limiting engagement with unwanted material. The team is now collecting data from teenagers aged 13 to 16 for a potential long-term study examining how algorithmic recommendations may affect emotional well-being over time.

More information: Carole Leung et al, Neural emotional processing of personally recommended short-video content and depressive symptoms, Computers in Human Behavior. DOI: 10.1016/j.chb.2026.109098

Journal information: Computers in Human Behavior Provided by University of Texas at Dallas