Generative AI Tools and Bilingual Language Learning: Student Practices and Pedagogical Implications

This project examines how undergraduate students use generative artificial intelligence (GenAI) tools such as ChatGPT, Gemini, Claude, and Copilot when completing academic work in a second language within a bilingual university environment. At the University of Ottawa, students study in a context where English and French coexist across programs and services. This setting offers a strong opportunity to explore how AI tools are used across languages.

Current discussions about GenAI in education often emphasize risks such as overreliance or academic integrity. There is still limited empirical evidence on how students use these tools in their everyday learning. This project addresses that gap through a close examination of student practices and their effects on language development.

The project has three main objectives:

1. Document how students use GenAI tools when completing academic tasks in a second language, including writing, revising, and reading.  
2. Evaluate how this use affects the quality of student language production, with attention to vocabulary, grammatical accuracy, and clarity.  
3. Identify effective and ineffective uses of GenAI in order to inform pedagogical practices in bilingual language education.  

The results will contribute to current discussions about GenAI in higher education and will be directly relevant to language teaching in bilingual institutions. 

The project follows a structured mixed-methods design carried out in several stages.

In the first stage, students complete a targeted literature review using a curated set of recent publications on GenAI in language learning and bilingual education. They produce annotated summaries and contribute to a shared synthesis of key findings.

In the second stage, students administer a survey designed in advance by the instructor. The survey gathers data on how undergraduate students use AI tools for language-related academic work, including frequency of use, types of tasks, and perceived benefits and limitations.

In the third stage, students conduct a controlled, task-based study. Participants complete short writing and revision tasks in either English or French using AI tools under defined conditions. Interactions with the GenAI systems, including prompts and outputs, are collected and anonymized.

In the fourth stage, students analyze the data. Survey results are processed using descriptive statistics. Task-based data are examined through qualitative coding with predefined categories such as error correction, paraphrasing, idea generation, and translation. Students compare initial drafts with AI-assisted versions to assess changes in linguistic quality.

In the final stage, students synthesize their findings and relate them to the literature. The analysis focuses on differences between English and French use and on how students move between the two languages when working with GenAI tools.

All activities follow established ethical guidelines. Students receive training in research ethics at the beginning of the project. They also receive structured guidance, templates, and regular feedback throughout.

• Students gain experience with the full research process, from reviewing literature to collecting, analyzing, and interpreting data. They develop skills in synthesizing academic sources, working with structured datasets, and applying qualitative coding methods. The project also introduces basic quantitative skills through the analysis of survey data.
• The project strengthens students’ ability to evaluate digital tools critically. Students examine how GenAI affects learning and language production and identify both strengths and limitations.
• Students also develop communication skills through written and oral presentation of their findings. The bilingual context encourages engagement with both English and French materials, and students may use both languages where appropriate. 

Of the 90 hours, 78 hours are allocated to independent research activities and 12 hours to meetings with the research team. This distribution falls within the expected range for the course.