Modelling Chat GPT for Language Learning - Prompting, Interaction, Best Practices

Starting from 2023-2024 Semester 2 at the Hong Kong Polytechnic University’s Centre for Independent Language Learning (CILL), I have been leading a series of workshops on teaching students how to responsibly use AI (ChatGPT) in language learning. This Teaching Demonstration will share how students have been trained to prompt ChatGPT for grammar and vocabulary feedback.
• Approaches to prompting – the concept of asking for feedback rather than asking for revision – this has been especially positively
received by HK PolyU students as they worry about being accused of plagiarism in using AI
• Approaching revision via communicative concerns, such as clarity, rather than simply asking about grammar mistakes
• The concept of using AI for language (grammar) choices rather than revising and rewriting, the concept of asking AI for choices
rather than re-writes
• Use relevant websites that also help with language learning, helping students take an eclectic approach to gathering feedback
rather than relying just on ChatGPT

Through this Teaching Demonstration attendees can gain ideas on how to teach students how to effectively and responsibly use AI. As
practitioners, this demonstration can spark dialogue with your department on how to form a pedagogy in ChatGPT instruction that works for
your students using modelling and best practices.

Theoretical Background

Research has shown that ESL students struggle with prompting ChatGPT and grasping the concept of using it as a tool, rather than having it write or revise for them (Klimova 2023). Some have theorised that the ed tech students have been accustomed to are mostly gap fill and multiple choice (Freiermuth and Zarrinabadi 2020); hence, experience in prompting and re-prompting could be less than we think. Many students have voluntarily attended this non-credit-bearing workshop to learn approaches and ways to prompt and work with ChatGPT to more skilfully use it as a tool. Instruction and examples are shared of adjusting, re-prompting, and evaluating usefulness of AI feedback.

After 3 semesters of running this, the feedback from students (both PG and UG) has been positive. The approach is driven by moulding behaviour in students rather than using policies, rules, and guidelines. Students are led through the process of thinking about what they want AI to do, reflecting on whether or not it is reasonable based on the literature (background and context). Once reasonable goals are established, the instructor takes those goals and models use of AI to meet those goals (Klimova 2023). Students see in real time the instructor prompting and further prompting, and making adjustments. Students are also given examples of past prompting and the results for further knowledge. Time is given for safe, monitored practice (Klimova, 2023) as the instructor is there to help if needed, and can answer questions.

Using AI can be a lonely, isolating endeavour. This step-by-step journey with the students also meets their need for relationships and shared experiences and gives positive moments. As noted in the Proceedings of the 20th International Conference of the Asia Association of Computer-Assisted Language Learning (2023), “The journey of language acquisition is not purely cognitive; there is a strong emotional dimension to it. While students navigate the highs and lows, anxieties and moments of exhilaration, a technological tool can supply answers,
but it lacks the depth and warmth of human interaction” (Pham et al, 2024, page 12); and the need for interaction and the teacher providing context and meaning (also in Klimova et al., 2023).

Hopefully this Teaching Demonstration can spark dialogue on how best we can teach and guide students on the best practices usage of ChatGPT.