Socio-Emotional Learning in AI K-12 Guidance and Policy Documents: A Gap Analysis

As generative AI becomes increasingly integrated into K–12 classrooms, it poses distinct socio-emotional risks that existing policies inadequately address. This study analyzed 36 institutional, state, and international AI guidance documents using the CASEL framework and five inductively generated SEL-informed categories: Anthropomorphization & Mental Model, Emotional Attachment & Parasocial Relationships, Engineered Manipulation & Supernormal Stimuli, Social Isolation & Relationship Displacement, and Developmental Vulnerability. Content analysis revealed uneven attention across categories: Developmental Vulnerability and Engineered Manipulation dominated, appearing in 88% of documents, while Emotional Attachment received minimal coverage (8%). The documents highlighted risks including diminished student agency, algorithmic bias, erosion of peer and teacher relationships, and the potential for parasocial engagement with AI. These findings reflect the need for policies that scaffold critical evaluation, maintain human oversight, promote relational and social-emotional development, and mitigate exploitation of developmental vulnerabilities.