AI-assisted GNW on Adolescent Emotional Distress

Adolescent mental health, and in particular emotional distress, has emerged as a global public health challenge. A recent meta-analysis of 80,879 adolescents found that approximately 25.2 percent experienced clinically significant depression and 20.5 percent suffered from anxiety, with these emotional disturbances accounting for an estimated 45 percent of the total disease burden in this age group. In China, the situation is equally alarming: the 2022 Blue Book on National Depression reports that roughly half of all individuals diagnosed with an emotional disorder are school-aged, and as many as 41 percent have interrupted their studies as a result. Beyond inflicting serious harm on adolescents' physical and psychological well-being, persistent emotional distress undermines social engagement and academic performance, and may precipitate more severe psychopathology, such as major depression, anxiety disorders, or even suicidal behaviors, placing a heavy burden on families, communities, and the public health system. Thus, early and effective intervention is of paramount importance for improving youth mental health.

Schools represent a central arena in adolescents' daily lives and an ideal platform for delivering psychological interventions. In response to rising rates of youth emotional problems, the Chinese government has enacted a series of policies to bolster mental health education and intervention in schools. However, the vast size of China's adolescent population and the relative scarcity of professional resources mean that traditional one-to-one intervention models have limited reach when implemented at scale. Many schools lack sufficient numbers of qualified counselors, and non-specialist teachers often do not possess the training needed to deliver sustained, targeted support, resulting in modest and non-durable effects on students' emotional well-being.

Moreover, school-based intervention programs confront several practical challenges. First, the number of licensed school-psychology professionals remains inadequate in many regions, leaving schools unable to provide consistent, high-quality support. Second, a persistent mismatch between intervention demand and available resources makes it difficult to tailor services to individual students' needs, undermining the overall effectiveness of existing programs.

Among the array of psychological interventions, writing-based approaches have gained prominence for their simplicity, low cost, and ease of dissemination in group settings such as classrooms. Writing interventions typically guide participants to articulate their emotions and thoughts in writing for 20-30 minutes per day over 3-5 consecutive days, thereby fostering emotional processing, adaptive expression, and the construction of personal meaning. Meta-analytic evidence confirms that such interventions yield significant improvements in emotional symptoms and related psychophysiological outcomes, including sleep quality. In Chinese adolescent populations, guided narrative writing has been shown to reduce examination anxiety and to ameliorate symptoms of depression and anxiety. Nevertheless, conventional writing interventions often lack mechanisms for timely, individualized feedback, which can dampen participant engagement and undermine the longevity of treatment gains.

Against this backdrop, the integration of artificial intelligence (AI) offers promising new avenues for enhancing psychological interventions. Prior research indicates that interventions incorporating ongoing, process-oriented feedback and interactive dialogue outperform those that provide no feedback. Yet, in large-scale implementations, human-delivered feedback is constrained by finite personnel resources. By leveraging natural-language processing techniques, AI systems can automatically detect emotional and cognitive markers in written text and deliver personalized feedback at scale, thereby addressing a critical gap in current practice. Introducing AI-driven feedback into adolescent writing interventions thus holds considerable promise for improving outcomes in emotional distress.

Building on these insights, the present study will recruit approximately 450 adolescent students and-using classes as clusters-randomly assign them in equal proportions to one of three arms: an AI-feedback Guided Narrative Writing group (AI-GNW), a no-feedback Guided Narrative Writing group (NF-GNW), or a Neutral Writing Group (NWG). We will evaluate the impact of AI-assisted narrative writing on emotional distress, comparing the trajectories of change across these three conditions.