Psychologists at the University of Texas at Dallas have found that social media algorithms may worsen users' mental health by serving personalized content that reinforces negative mood. The researchers published their findings in a study examining the links between algorithm-recommended content, brain activity and depressive symptoms.
The study assessed 60 young adults with an average age of 20 using validated anxiety and depression scales. Participants watched personalized videos from their own Instagram or TikTok accounts as well as generalized trending videos while researchers measured their brain activity using EEG to assess emotional processing.
The study found that participants reporting more depressive symptoms were shown recommendations containing more depressive content. Their brain activity, measured through frontal alpha asymmetry, showed patterns characteristic of depression when watching personalized feeds—but not when viewing trending videos. Dr. Stacie Warren, an associate professor and study co-author, said the immediate neural measurement provided an objective way to evaluate engagement with algorithm-selected content.
Social relationships, particularly romantic or friendship-related content, made up the most common category of recommended videos. The researchers found that depressed individuals who continued scrolling encountered fewer positive videos, a pattern associated with more negative feelings. Dr. Alva Tang, the study's corresponding author, said the research differed from previous work relying on self-reports by measuring real-time emotional processing as users watched recommendations.
The researchers suggested that people could manage their feeds more actively by following specific creators, resetting accounts to defaults, and curating positive content. Tang noted that educating young people on effective engagement with social media may be more practical than simply limiting screen time. The team is now collecting data on adolescents ages 13–16 to study the longer-term emotional effects of algorithm recommendations.