Task Design with AI to Promote Engagement with Feedback: A Case Study

By Linh Phung, EdD

Author's bio: Dr. Linh Phung is the Founder of Eduling, an application and technology platform that offers English and teacher development courses as well as content creation tools for the development of communicative tasks and games. As a researcher, she has papers published in high impact journals, including Language Teaching Research and Studies in Second Language Acquisition. Her professional experience also involves working as the Director of the English Language Program at Chatham University for 12 years and serving as an English Language Specialist with the U.S. Department of State. As an author, she has published several language learning and children’s books.

Slides of the presentation at TESOL 2026 can be accessed HERE.

Introduction

In the current academic landscape, it is increasingly common for students and academic writers to utilize Generative AI to draft their work. While there is concern that learners may over-rely on AI to write for them and present the output as their own, the demands of standardized assessments, such as the IELTS exam, which requires the independent production of two essays within 60 minutes, can encourage learners to use AI tools to build their own skills. By capitalizing on this motivation, educators can integrate specially designed AI-mediated tasks into instruction to scaffold the learning process and support students in achieving their target writing goals.

This article reports a case study of five learners from Vietnam who completed an interactive practice task with EdulingAI by writing sentences and paragraphs to discuss the advantages and disadvantages of knowledge available on the Internet. The article focuses AI-mediated task design on a mobile app to develop second language writing and its potential for autonomous and self-regulated learning. By breaking a writing project into smaller tasks, specific steps such as brainstorming, drafting, and revising based on feedback can be guided by AI. This approach has the potential to not only promote language development and improve writing quality but also enhance self-regulated learning.

Framework for Self-Regulated Learning

Drawing from Reinders et al.’s (2023) systematic framework, self-regulated learning in writing involves several key components: motivation, identification of learning needs, goal setting, planning, and task regulation. Central to this process is self-assessment. Task regulation is defined by the selection of resources, application of strategies, and monitoring of progress. Within this framework, the main focus of this article is on the concept of student engagement with feedback (Zhang and Hyland, 2022) as a site for learning because it requires learners’ focus on both meaning and language and active reflection and action.

Method

Five learners from Vietnam, who were preparing for the IELTS exam, volunteered to complete a task with EdulingAI on the Eduling app and answer interview questions after task completion through Zoom in February 2026. The participants, who self-assessed their English proficiency at IELTS bands 5.5–7.0, completed a task regarding the pros and cons of the Internet while looking at a picture with some suggested ideas for the topic (Figure 1). Discussing the advantages and disadvantages is a common essay question in the IELTS exam. 

EdulingAI is a specially designed dialogue agent for task-based language teaching (TBLT). It enables multi-modal task-based interactions with the learner to achieve task goals. 

Figure 1: Pros and Cons of Internet Task on Eduling

The task was structured to break the writing process into manageable segments rather than a single 250-word essay. Students were prompted to write an introduction, followed by individual sections for advantages and disadvantages, and a conclusion. Throughout the process, the AI prompted students to elaborate on their ideas and provided various forms of feedback, including recasts, prompts, and explicit corrections.

The five students spent 15-30 minutes completing the task. Some chose to type their responses while others recorded them orally in the interaction with EdulingAI. 

Finding 1: The learning opportunities from the AI

Based on the screenshots shared by the learners, the AI provided several types of learning opportunities, including prompting learners to elaborate on their ideas, negotiating meaning when part of their response was not clear, flexibly offering corrective feedback moves (recasts, elicitations, and corrections), and providing an overall assessment of their writing. These features are not common in generic chatbots but were specially designed by Eduling to promote second language acquisition. The results discussed below were obtained using EdulingAI. Consumer AI models may not necessarily produce similar results.

Below are examples from this study.

Excerpt 1: Example of negotiation of meaning 

Learner: The development of Internet brings both good and bad sides to people. So we should erase carefully. 

EdulingAI: What do you mean by “erase carefully?” 

Learner: Use it 

EdulingAI: Ah, you mean “use it carefully.” 

Excerpt 2: Example of explicit correction 

Learner: On the other hand, the internet brings thing call cyberbullying to people. For instance, many youngsters now would like to post their photos on their biography and this is the chance for bad people to blame them. 

EdulingAI: We say, “the Internet brings something called cyberbullying to people.” 

Excerpt 3: Overall feedback 

EdulingAI: Clarity: “Isolation of users on social” could be “social isolation of users." 

Grammar: “Lead to the dependence” should be “leading to the dependence.” 

Structure: Consider breaking the sentence for clarity. 

Finding 2: Favorable aspects of the AI-mediated task

Participants identified several positive aspects of the AI-mediated task. The structured step-by-step design of the interaction, which involved the production of a few sentences at a time, reduced cognitive load. Learners found the task much more manageable and less overwhelming. One learner said, “I really like that I can break the essay into small steps, and then the AI goes through those small steps with me one by one. I feel less overwhelmed than when writing the whole essay at once.”

Feedback was provided within seconds and focused on vocabulary, spelling, and grammar, which the learners noticed and commented on. They also appreciated “vocabulary upgrades” that the AI suggested such as “purchase” instead of “buy.” In addition, the AI’s prompts encouraged students to provide more detailed examples by asking questions and giving suggestions.

Finding 3: Suggested features and improvements

Learners also proposed several improvements for the AI-mediated task and interaction, including an additional stage to rewrite sentences after reviewing feedback from the AI. This reflects their interest in more elaborate engagement with the task by having a chance to reformulate their output. There was a desire for the AI to suggest more complex ideas to achieve higher target scores. 

In addition, as the IELTS exam scores examinees on a 0-9 scale, learners wanted to set specific target band scores to receive tailored feedback from the AI. “At the beginning of the chat, if it could ask me about my target band score and then give feedback based on that, I think it would help me improve more,” said one of the participants. Finally, another suggested a built-in dictionary or word look-up tools, indicating students’ heightened attention to form and desire for an integrated tool that meets their various learning needs.

Finding 4: Impact on autonomous learning

When asked about the potential of completing AI-mediated tasks on a mobile app for autonomous learning, students highlighted the convenience and accessibility of studying on the go without needing a laptop. Many also noted the reduced research time required to brainstorm ideas for various test topics. This benefit was captured in one student's comment: “The time to research ideas for topics I haven't known much about is a lot, so if we can practice like this immediately, it makes sense.”

Beyond time savings, the segmented nature of the tasks made the writing process feel more manageable and less overwhelming. Furthermore, explicit feedback allowed students to identify and correct their own errors immediately, rather than waiting for a teacher's response. Collectively, these factors, i.e. convenience, immediate feedback, and reduced cognitive load, lowered barriers to autonomous learning and active engagement, both of which are essential for long-term language development.

Conclusion

In conclusion, the interactive design of the task on the Eduling app allows learners to work directly with AI while utilizing their own linguistic resources. The students in this case study did not rely on AI to write for them; instead, they produced their own language and engaged actively with the feedback provided. By breaking the writing process into manageable segments, this approach lowers cognitive barriers and reduces the time-consuming research typically associated with unfamiliar topics. 

By aligning task design and technological affordances with student motivation, particularly the need to develop skills for independent exams, educators can promote active and beneficial engagement with learning activities and technology. As noted by Marta Gonzalez-Lloret in a March 2026 interview on The Language Innovators Podcast, the rapid development of these tools offers significant possibilities for autonomous learning. "Then why not?"

NOTE: Try the task discussed in this article HERE. Clicking the link will take you to the App Store or Play Store to download the app and then take you to the task.

References

Reinders, H. (author), Phung, L. (consultant), Ryan, S. (consultant), Thomas, N. (consultant), (2023). The Key to Self-regulated Learning – A systematic approach to maximising its potential [PDF]. Oxford University Press. www.oup.com/elt/expert

Zhang, Z., & Hyland, K. (2022). Fostering student engagement with feedback: An integrated approach. Assessing Writing, 51, Article 100586. https://doi.org/10.1016/j.asw.2021.100586