Comparing Children’s Verbal Engagement with an Interactive Media Character and a Human during Video Watching (Stage 3, 9:59 AM)

• In Event: e-Lightning Ed-Talk Session 16 - Stage 3

Abstract

This study examined how children’s verbal engagement differs when interacting with an AI-powered conversational agent versus a human co-viewer during video watching. 77 children aged (4-7) participated, engaging in question-and-answer interactions with an AI-powered conversational agent or a human partner. Five aspects of verbal engagement were measured: response rate, response length, lexical diversity, response accuracy, and response relevance. Regression analysis controlling for age, gender, language proficiency, and prior science knowledge showed children interacting with the AI agent produced shorter responses and marginally lower lexical diversity. Response rate, accuracy, and relevance did not differ by condition. Language proficiency predicted higher verbal engagement across conditions. Findings suggest human partners may better elicit elaborative language, highlighting the need for more responsive conversational agents.

Authors

• Nikki Nasseri, University of California - Irvine

• Kunlei He, University of California - Irvine

• Mark Warschauer, University of California - Irvine