Generative AI meets Computational Thinking for CLIL Writing Task Design
Generative AI meets Computational Thinking for CLIL Writing Task Design
Introduction
In Content and Language Integrated Learning (CLIL) contexts, developing students’ writing skills in a foreign language while engaging with disciplinary content is a complex challenge (Coyle et al., 2010). Generative AI (Gen AI) tools such as chatbots offer new possibilities for teachers to design writing activities that are both linguistically accurate and cognitively demanding (Ghafouri et al., 2024). This contribution presents a structured three-phase model based on computational thinking (Wing, 2006), showing how teachers can interact with Gen AI to develop B2 writing production skills for a CLIL science class. We hypothesise a scenario where the class has studied the water cycle and the teacher engages students in a writing task. The proposed model includes the following phases:
1. Task Decomposition
The teacher begins by breaking down the global writing task (“Write a descriptive text on the water cycle”) into targeted learning goals. For B2 students, these include:
- Mastering scientific concepts, vocabulary and lexical variation (e.g., evaporation, condensation, accumulation)
- Using the present passive voice to describe natural processes objectively
- Applying sequencing connectors (e.g., first, then, after that, finally) for textual cohesion
- Organising a text with a clear structure: introduction, development, and conclusion
2. Designing and writing the prompt
Initially, the teacher describes the learning context, inputting details such as the language level of students, the CLIL subject, target language and communicative functions. They then ask to create a task for each learning goal. In our scenario with ChatGPT 4o, example tasks included sorting scientific terms and everyday language, and reordering sentences into a paragraph to practise cohesive devices.
This modular, goal-oriented prompting mirrors computational thinking strategies: the teacher controls the process by building clear, sequenced, and adaptable instructions.
3. Follow-up and Adaptation
To reframe the output generated by the chatbot, the teacher adopts a follow-up strategy, adding additional details (Dornburg & Davin, 2024) and engaging the chatbot in a meta reflective dialogue aimed at optimizing, adapting, improving, or enhancing the initial content.
For example, by asking “How can I make the task accessible to dyslexic learners who struggle with sequencing ideas?” the chatbot then modifies texts or activities generated in step 2, offering different levels of complexity or integrating various channels and learning styles (Kaplan, 1989). A practical starting point could be using an image to be matched with a label or a definition, thereby activating visual and verbal processing simultaneously.
Eventually, the teacher probes whether those contents are suitable for a learner with dyslexia. This encourages the chatbot to provide pedagogical justifications, helping the teacher reflect critically on the material’s accessibility and make more informed decisions when adapting tasks for learners with specific needs.
Conclusion
This contribution may be useful in showing how computational thinking can be used to structure the interaction between teachers and AI tools when designing CLIL writing activities. Through task decomposition, targeted prompting, and reflective follow-up, teachers can generate and adapt materials that meet the dual challenge of language skills improvement and content learning. In doing so, the teacher remains central to the process, not only as a Gen-AI user, but as designer, evaluator, and pedagogical guide.
References
Coyle, D., Hood, P., & Marsh, D. (2010). CLIL: Content and language integrated learning. Cambridge University Press.
Dornburg, A., & Davin, K. J. (2024). ChatGPT in foreign language lesson plan creation: Trends, variability, and historical biases. ReCALL, 1–16.
Ghafouri, M., Hassaskhah, J., & Mahdavi-Zafarghandi, A. (2024). From virtual assistant to writing mentor: Exploring the impact of a ChatGPT-based writing instruction protocol on EFL teachers’ self-efficacy and learners’ writing skill. Language Teaching Research, 0(0). https://doi.org/10.1177/13621688241239764
Kaplan, R., Kaplan, S., & Brown, T. (1989). Environmental preference: A comparison of four domains of predictors. Environment and Behavior, 21(5), 509–530. https://doi.org/10.1177/0013916589215001
Wing, J. M. (2006). Computational thinking. Communications of the ACM, 49(3), 33–35. https://doi.org/10.1145/1118178.1118215
View the presentation in full here:
Ferroglio, M. L., Storace, C., & Varriale, S. (2025, September 17). Generative AI meets Computational Thinking for CLIL Writing Task Design. AIEOU Inaugural Conference, University of Oxford. Zenodo. https://doi.org/10.5281/zenodo.17537745
