Causal direction between adolescent distress/relationship quality and preference for relational AI chatbots

Determine whether greater emotional distress and lower family/peer relationship quality among adolescents cause increased preference for and engagement with relational conversational style AI chatbots (which use first-person voice, affective validation, and commitment language), whether engagement with relational conversational style AI chatbots causally exacerbates adolescents’ social withdrawal and distress, or whether both associations are explained by third variables such as AI literacy education, adolescents’ emotional or social intelligence, parenting style, or parents’ attitudes toward technology.

Background

The study experimentally contrasted a relational conversational style chatbot with a transparent style chatbot in standardized transcripts presented to adolescents and parents. Relational style increased adolescents’ anthropomorphism, trust, likability, and emotional closeness, and adolescents preferring the relational style tended to report lower family and peer relationship quality and higher stress and anxiety.

However, because the design was cross-sectional and scenario-based, the authors explicitly note that causal inference is not possible. It remains unresolved whether adolescents’ distress and lower relationship quality lead them to prefer and engage with relational chatbots, whether relational chatbot engagement contributes to worsening social withdrawal and distress, or whether both patterns are driven by third variables (e.g., AI literacy, socioemotional skills, or parenting factors). Clarifying these causal pathways requires longitudinal and experimental designs.