The Anatomy of "Sticky Chats": Identifying and Disrupting Five Addictive Interaction Patterns in Child-AI Conversations

• In Event: AERA Poster Session Sunday 11:45 am
  In Poster Session: Early Education and Child Development

Abstract

As child-facing AI characters become increasingly conversational, concerns emerge about "sticky chats" referred to as interaction patterns prolonging engagement without supporting reflection or exit. Through analysis of survey transcripts from 30 families with young children (ages 5-8), we identified five addictive interaction patterns in child-AI conversations: (1) looping prompts, (2) simulated friendships, (3) pseudo-praise, (4) convincing hallucinations, and (5) false promises. These patterns limit children's agency and self-regulation while increasing emotional dependency on AI characters. We advocate redesigning AI characters to disrupt these patterns through five targeted approaches: (1) clear closure, (2) shared role awareness, (3) fair assessment, (4) acknowledging lack of information, and (5) human redirection.

Authors

• Sonia Tiwari, Kids AI

• Jennifer J. Chen, Kean University