Sentiment Changes During Anonymous, On-Demand, Real-Time Moderated Digital Emotional Support Among Government Plan-Sponsored Adolescents and Young Adults: A Retrospective Observational Study

Abstract

Background. Adolescents from low socioeconomic status households experience increased emotional distress and barriers to mental health care. Real-time moderated digital emotional support (MDES) may offer a scalable and low-barrier complement to formal services. Objectives. This study focused on state-level sentiment changes among government plan-sponsored adolescents and young adults who accessed a no-cost, 24/7, anonymous, real-time MDES service. Methods. The retrospective analysis included 1005 moderated live chats between March and December 2024 from 663 users aged 13–20. Sentiments were scored using a few-shot learning approach with GPT-4o-mini, evaluating sadness, stress, insecurity, loneliness, anger, anxiety, and optimism. Sentiment trajectories were mapped as conversations progressed and assessed using Wilcoxon signed-rank tests and effect sizes. Results. Significant improvements were observed in all sentiments throughout the conversations (all p < 0.001), with the greatest reductions seen for anger (46.94%, r = 0.87), loneliness (31.99%, r = 0.79), and stress (31.02%, r = 0.82). Smaller but also significant reductions in anxiety (29.81%, r = 0.79), sadness (28.64%, r = 0.81), and insecurity (24.17%, r = 0.75) were observed, with a substantial improvement in optimism (47.69%, r = 0.70). Conclusions. Among government plan-sponsored adolescents and young adults, use of anonymous, real-time MDES was associated with reduced negative sentiment and increased optimism expression. Given the higher burden of mental health concerns and barriers in this population, this service may serve as a scalable approach to expand access to emotional support.