The effect of immersive conversational AI on immigrant students' second language production and willingness to communicate

Recognizing the challenges that immigrant Latino youth in the U.S. encounter, there is a growing interest in harnessing Artificial Intelligence (AI) for English as a Second Language (ESL) learning (Ayedoun et al., 2019). While existing research predominantly involves university-level participants (Woo & Choi, 2021) and focuses on students' perceptions (Jeon & Choe, 2023), there is a need to investigate the potential of conversational AI for ESL in a naturalistic high-school setting (Katsarou et al., 2023). Moreover, there has been limited research on immersive conversational AI, which combines the benefits of Virtual Reality (VR) in reducing the affective filter (Thrasher, 2022) with AI's potential for personalized learning experiences (Huang et al., 2023). Responding to the call for research on immersive conversational AI, this mixed-methods study, guided by the ecological perspective and Complex Dynamic Systems Theory (CDST) (Larsen-Freeman, 2018), explores the impact of immersive conversational AI on language production and willingness to communicate (WTC) in an underserved Illinois high school. Over one academic quarter, thirty-six (N = 36) immigrant Latino high school students in ESL Resources class engaged in 20-minute-long weekly role-plays with AI-powered bots through the Immerse application, which was first accessed on desktop and subsequently in VR. Interactions with AI bots were recorded and analyzed for students' language production and usage of conversational features. Students' WTC was also evaluated through a questionnaire by Dewaele and Dewaele (2017). During data analysis, a Spearman correlation test was conducted to explore relationships between ESL level, immersion level, modality preference, WTC, language production, and specific conversational features. Findings from this study indicate that lower-level ESL students preferred desktop interactions and demonstrated lower initial willingness to communicate, but they exhibited significant growth in WTC compared to advanced ESL peers, who favored VR and started with higher WTC levels. Beginner and low-intermediate students using desktop showed greater growth in language production. While lower-level students engaged more frequently with all conversational features, only the word pronunciation feature positively correlated with higher language production. Additionally, two disruptions in human-AI communication patterns were identified: limited recognition of mispronounced words, impacting lower-level ESL students, and issues with recognizing contextual cues across all levels. These findings underscore conversational AI's potential to enhance language production and willingness to communicate among lower-level immigrant ESL students, supporting its integration into high school curricula.

Name thesis:degree_name

M.A.

Level thesis:degree_level

Thesis

Discipline thesis:degree_discipline

Teaching of English Sec Lang

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2024

Author dc:creator

• Ovsiannikova, Uliana

Contributors dc:contributor

• Sadler, Randall
• Thrasher, Tricia
• Mroz, Aurore
• Chun, Dorothy

• Immersive Conversational Ai
• Virtual Reality For Language Learning
• Ai Role-plays For Language Learning
• Willingness To Communicate
• Language Production
• Immigrant Students
• Underserved High School

Statement dc:rights

• Copyright 2024 Uliana Ovsiannikova

Language dc:language

en, eng

Handle dc:identifier

https://hdl.handle.net/2142/125591

OAI record oai:identifier

https://www.ideals.illinois.edu/oai-pmh?verb=GetRecord&metadataPrefix=etdms&identifier=oai%3Awww.ideals.illinois.edu%3A2142%2F125591