Design and evaluation of an AI-assisted intercultural communication tutor for second language learners

Thesis Abstract

This study addresses the growing need for scalable, personalized intercultural competence development in second language (L2) education by designing and evaluating an AI-assisted intercultural communication tutor (AI-IC Tutor) that integrates language proficiency with intercultural awareness and pragmatic skills. The problem is the inadequate real-time feedback on intercultural nuances and pragmatic appropriateness in conventional L2 learning environments, which often neglect contextual and cultural variability across communicative situations. The aim is to develop a responsive AI tutor that adaptivey scaffolds intercultural communication tasks and evaluates learner performance against culturally informed benchmarks. Specific objectives are (1) to design an AI-IC Tutor capable of generating context-rich dialogues, feedback grounded in intercultural communication theories, and adaptive difficulty based on learner trajectories; (2) to implement a multimodal data collection framework comprising written, spoken, and interactional traces to assess pragmatic appropriateness, discourse management, and cultural sensitivity; (3) to evaluate the effectiveness of the AI tutor in improving intercultural communicative competence (ICC) and L2 proficiency relative to a control condition; (4) to examine learner engagement, perceived usefulness, and autonomy in using the AI tutor; and (5) to validate a theoretical model linking AI-mediated feedback to ICC development. The study employs a mixed-methods, quasi-experimental design within higher education institutions in a multicultural city. The population includes undergraduate L2 learners enrolled in English as a Foreign Language courses, with a target sample of 240 participants aged 18–25, randomly assigned to an experimental group (n=120) using the AI-IC Tutor and a control group (n=120) receiving standard instruction over a 12-week intervention. Data collection instruments comprise (i) standardized ICC measures (e.g., the ICCS—Intercultural Communication Competence Scale) administered at pre-, mid-, and post-test; (ii) L2 proficiency assessments (speaking and writing tests aligned with CEFR benchmarks); (iii) interaction datasets from AI logs capturing dialogue acts, pragmatics, and feedback reception; (iv) validated perception surveys measuring perceived usefulness, ease of use, and autonomy; and (v) semi-structured interviews with a purposive subsample (n=40) to explore experiences and perceived cultural sensitivity in AI feedback. Quantitative data will be analyzed using mixed-effects regression to assess changes in ICC and L2 proficiency over time, with group, time, and interaction effects. Mediation analyses will examine whether improvements in ICC mediate the relationship between AI use and communicative outcomes. Qualitative data will be analyzed thematically to identify patterns in learner perception, cultural alignment of feedback, and perceived realism of intercultural scenarios. The AI tutor will be grounded in two complementary theories intercultural communicative competence theory (Bennett, 1993) and sociocultural theory of learning (Vygotsky), with the feedback mechanism operationalized through Scaffolding and Zone of Proximal Development (ZPD) constructs. Expected findings include significant gains in ICC and pragmatic competence for the experimental group, higher engagement and perceived usefulness, and positive perceptions of feedback relevance and cultural sensitivity. The study anticipates that AI-mediated, context-aware feedback will outperform conventional instruction in promoting nuanced discourse management, appropriate language use across intercultural contexts, and reflective awareness of cultural norms. Theoretical contributions involve integrating ICC theory with AI-driven pedagogical scaffolding models, advancing understanding of how adaptive feedback shapes intercultural learning in L2 contexts. Practically, the research will provide a scalable blueprint for deploying AI-assisted intercultural tutors in language programs, including design principles for culturally diverse corpora, safety and bias considerations, and data governance. Policy implications include recommendations for curriculum integration, teacher professional development, and assessment frameworks recognizing intercultural competence as an essential dimension of language proficiency. The study concludes that the AI-IC Tutor is an effective tool for enhancing ICC and L2 outcomes, with recommendations to extend the platform to additional languages, incorporate peer collaboration features, and conduct longitudinal follow-ups to examine transfer to real-world intercultural encounters.