Impact of Conversational Agent Language and Text Structure on Student Language

This study examines how conversational agents’ language (formal vs. informal) and text structures (comparison vs. causation) impact student language in written summaries using an intelligent tutoring system (ITS). We used the Coh-Metrix-ENA approach, which integrated Coh-Metrix and epistemic network analysis (ENA), to examine the structure of language connections in students’ written summaries. Results revealed both agent language and text structures impacted student language. Specifically, informal agent language elicited a stronger language connection between nonnarrativity and word abstractness in both causation and comparison texts in the posttest network. However, in causation texts, both informal and formal agent language induced a weaker language connection between syntactic complexity and deep cohesion, which improved students’ tendency to overuse causal connectives. Moreover, in comparison texts, informal agent language elicited a weaker language association between syntactic complexity and referential cohesion. Implications are discussed regarding the design of conversational agent language and text characteristics for interventions to improve summarization strategies and academic language skills.