Adolescent AI Use and Developmental Needs: A Self-Determination Theory Perspective

• In Event: AI and Youth Learning Across the Globe: Insights from Four Large-Scale Surveys

Abstract

Objective
Adolescents are increasingly incorporating generative artificial intelligence (AI) into everyday activities (Madden et al., 2024), highlighting the need to understand how specific forms of AI use may influence psychological needs that are central to adolescent well-being. This study examined the landscape of teen AI use, identified clusters of activities, and assessed how each predicts adolescents’ perceptions of generative AI’s impact on their competence, autonomy, and relatedness.

Theoretical Framework
According to Self‑Determination Theory (SDT, Deci & Ryan, 2012), psychological well-being depends on fulfilling needs for competence (feeling effective), autonomy (acting with volition), and relatedness (feeling socially connected). Uses and Gratifications Theory (UGT, Ruggiero, 2000) complements this by suggesting that individuals turn to media to meet specific instrumental and expressive goals.

Methods
We conducted a national online survey in August 2024 using the Alchemer platform. The sample included 1,440 U.S. adolescents aged 13–17 (M = 15; 49.2% girls; 50.8% white), with quotas set to approximate U.S. Census benchmarks. Participants reported how often they used generative AI for 14 activities (e.g., essay writing, image generation, seeking advice). They also rated AI’s perceived impact on seven ability areas mapped to SDT domains: task performance and attention (competence), curiosity, imagination, hobbies, and self-directed thinking (autonomy), and social skills (relatedness).

Results
Among our sample, 53.7% had used generative AI, 32.9% were aware but non‑users, and 13.4% were unaware of AI tools. For the 790 users, exploratory factor analysis (EFA) revealed a three‑factor structure accounting for 62% of variance (χ²(52)=177.2, p <.001; RMSEA=.055 [.046-.064]; TLI=.968; RMSR=.02). Teens most frequently reported Homework-Related Use, followed by Creative Use and Personal Use (Figure 1).
Multiple regression analysis (Figure 2) revealed Homework-Related Use was the strongest predictor of perceived competence impact (β = .29, p < .001, R² = .16), followed by Creative Use (β = .15, p < .001), while Personal Use was not significant (β = .07, p = .06). Autonomy impact was most strongly predicted by Creative (β = .18, p < .001) and Personal Use (β = .16, p < .001), with a smaller contribution from Homework-Related Use (β = .11, p < .001). Relatedness impact was primarily predicted by Personal Use (β = .28, p < .001), followed by Creative (β = .14, p = .006), with Homework-Related Use contributing modestly (β = .09, p = .042). Age was negatively associated with all outcomes; gender was not significant.

Significance
This study offers a developmental framework for understanding how adolescents’ AI use relates to their perceptions of AI's impact on core psychological needs. Findings suggest that AI tools designed for academic support may reinforce competence, while those fostering exploration and emotional expression may better serve autonomy and relatedness. Future design and policy efforts should explicitly consider how specific AI features, such as guided prompts, collaborative creativity, or conversational feedback, can be aligned with the developmental needs of adolescents across diverse contexts.

Authors

• Zhiying Yue, Harvard University

• Hannah Chidekel, Harvard University

• Michael Rich, Harvard University

• David S Bickham, Boston Children's Hospital