Grow with Your AI Buddy: Designing an LLMs-based Conversational Agent for the Measurement and Cultivation of Children's Mental Resilience

Psychological resilience refers to an individual's ability to adapt to adversity and stress. Education on psychological resilience during childhood can contribute to future mental health and well-being, such as reducing anxiety and depression [1] [2]. However, traditional psychosocial resilience training faces challenges with accessibility, heavily constrained by cost and spatiotemporal limitations. Recently, emerging large language models (LLMs) have demonstrated exceptional capabilities in conversational tasks, indicating new prospects for cultivating children's psychological resilience. In our work, 1) we conducted qualitative interviews with 10 Chinese children (aged 8-12) and their parents to understand their needs and current conditions; 2) based on the interview results and theories of psychological resilience, we summarized three pathways for developing children's psychological resilience using conversational agents (CAs) and identified six key challenges for designing child-centered CAs; 3) we designed and developed a web prototype using optimized LLMs (see Figure 1), which integrates personal and social support factors, to measure and foster children's psychological resilience through conversations; and 4) we invited 48 child volunteers in user testing and designed three sets of experiments to evaluate the effectiveness of system interventions, the effectiveness of measurements, and overall acceptability. Results indicate that the intervention tasks actively promoted psychological resilience in adolescents. Intelligent measurement scores were effectively consistent with traditional scales in objective scoring, while subjective evaluations, such as appeal and fun, significantly exceeded traditional scale scores.