Background: To support participatory research about social media (SM) and mental health, we co-designed SocialsVoice with community stakeholders—a Photovoice adaptation using SM, theory, and a convergent mixed methods randomized design. Prior studies have excluded qualitative and SM narratives from youth and instead relied on survey or smartphone data, limiting trust and impact.

Method: In 2025, youth ages 13-24 (N=41) from U.S. community-based organizations were randomly assigned to share SM clips showing mental health positive or negative content as defined by youth. In 7 video-chat group sessions, youth discussed clips, authored captions, screened co-created summary videos, and completed pre-posttests. Data from SM and qualitative codes, were triangulated with ANOVA statistics to explore participatory effects and meta-inferences on mental health symptoms, self-perceptions, stigma, and SM use behaviors.

Results: The negative vs positive theme group improved CES-DC and PHQ-4 scores (p<.05). In contrast, the positive vs negative group reported improved self-rated mental health (p<.01). Mental illness knowledge/attitudes declined in both groups while social distance improved in the negative group only; these trends were not significant. The negative vs positive group reduced neurobiological and increased psychosocial causal attributions (p<.10). Lastly the negative vs positive group reduced SM problem use, social contact, peer comparison, and body/image concerns, though not significant. The presentation will share joint display tables of mean scores with SM and qualitative excerpts to form meta-inferences (confirmation, discordance, expansion).

Conclusion: SocialsVoice goes beyond typical data science by power-sharing to co-create an integrated database of real-world SM experiences documenting both the harms and benefits of SM for mental health. To advance pragmatic policy about SM safety and utility, a free educational book on SocialsVoice findings was sent to participants, policymakers and key stakeholders.