Section A: Project Information
The project targets to give feedback to Primary and Secondary students who are learning English as their second language (L2). It involves an AI writing feedback bot “Writing-helper-L” to provide written feedback and a platform for teachers to record audio-feedback for students. On one hand, the project developed a prototype of AI writing assistant that could provide written feedback on the students’ writing assignments. After the users typed in a sentence or a passage, it starts by listing the local grammatical errors found in the writing. Next, the AI will suggest improvements or corrections with explanations considering the level of the learners. The AI will also provide overall feedback on the global level errors of the writing. Lastly, it will provide a corrected version of the user's writing passage.
On the other hand, the project suggests introducing an audio-feedback system on Padlet to English Language Teaching (ELT) school teachers. The teachers can upload students’ writing assignments to the system on Padlet. Then the teachers can record audio-tracks to comment on the excellent and critiqued parts of the writing to students. Unlike AI feedback assistant, teachers are given guidelines and suggestions on giving individualised feedback. The audio-feedback could facilitate teachers to give more individualised feedback with Feed-up (what students can be added to the same task), Feedback (comment on writing performance presented in the task) and Feed-forward (give suggestions on specific directions for subsequent tasks). The students could listen to the feedback provided by their English teachers, who have a great familiarity with the learners’ ability, improvements and skills.
Combining the advantages of the new AI writing assistant and Audio-feedback system, students can gain comprehensive feedback and make progressive improvements. Both designs are aiming at enhancing students’ writing skills and the efficacy of using audio feedback to improve teaching effectiveness.
Section B: Participant Information
| Title | First Name | Last Name | Organisation/Institution | Faculty/Department/Unit | Phone Number | Current Study Programme | Current Year of Study | Contact Person / Team Leader | |
|---|---|---|---|---|---|---|---|---|---|
| Miss. | Kiu Lydia | Lee | Hong Kong Baptist University | Department of Education and Psychology | lydialeekiu@gmail.com | 53088100 | Master's Programme | Year 5 |
Section C: Project Details
A common issue in English Language Teaching (ELT) classrooms in Hong Kong is that teachers rarely provide individualized comments or feedback on students' writing assignments, which hinders future improvement. One reason for this is that in-service ELT teachers often lack the time to give detailed, one-on-one feedback after returning assignments. Instead, they tend to point out only common errors due to constraints on class time. Additionally, many teachers find it challenging to reference specific sections of students' handwritten work for feedback. As a result, teachers typically only mark local errors, assign grades within each marking domain, and fail to offer significant individualized feedback for overall improvement. Research has shown that audio feedback can benefit students' language learning and enhance their writing performance across various educational levels, from primary school to university (Erkan, 2022; Ab Hamid & Romly, 2020; Cavanaugh, 2014; Silva, 2012). Receiving constructive feedback on assessments is crucial for learners to make necessary corrections and improvements. By providing more individualized comments, teachers can better address students' strengths and weaknesses.
The potential solution for solving the problem of insufficient individualized comments found in students’ writing assignments is to introduce AI writing feedback assistant and the audio-feedback system in ELT to in-service teachers in Hong Kong, and guide the teachers in using audio-feedback in students’ formative writing assignments. This project suggests that English teachers ask their students to do Process Writing for the writing assignments. Divide the post-writing activities into Part A, B and C. In Part A for the 1st Draft, ask students to write the article by themselves and check their writing. Students can do self-evaluation with “My Writing Ruler”. For Part B for completing the 2nd Draft, ask their students to use the AI writing feedback assistant “Writing-helper-L” to check the grammar, language, fluency, clarity, and writing structure errors. In Part C for making the 3rd Draft, students can listen to the Human Teacher’s audio-feedback and make final changes to their writing.
n/a : stream 3
I started designing the Audio-feedback system on Padlet at the beginning of March, dedicating approximately three weeks. Following that, I spent one week designing and modifying the “Writing-helper-L” on Poe, making several adjustments along the way. The base bot of “Writing-helper-L” is the “Claude-3-Haiku” on Poe. My AI bot is equipped with documents retrieved from a grammar book and guidelines designed by me on how to provide individualised feedback for students' writing.
My AI writing feedback bot starts by identifying local grammatical errors in the writing, including punctuation, word choice, spelling, grammatical mistakes, sentence structure and cohesion and so on. Next, the AI will suggest improvements or corrections with explanations considering the level of the learners. The AI will also provide overall feedback on the global-level errors in the writing. Lastly, it will provide a corrected version of the user's writing passage. Following local errors, the AI will provide 🔍Overall feedback on the global level errors of the writing, such as asking questions for amplifying, reorganizing the paragraph structures, sharpening the focus, and changing the tone or formality of the text. It will provide a corrected version of the user's writing passage at the end. I would adopt the following performance metrics to evaluate the effectiveness: accuracy, comprehensiveness, and presence of details.
The same guidelines and suggestions on giving individualized feedback are given to the English in-service teachers. First, the teachers ask the students to upload their writings (1st draft) and the feedback from the AI writing feedback assistant (2nd draft). Next, upload their marked students’ writing and record audio-feedback to comment on the writing. To evaluate the effectiveness of the audio-feedback, I would use the following performance metrics: variety of feedback types, comprehensiveness, and presence of details.
The existing AI writing tools primarily focus on grammar checks by correcting local-level errors, such as punctuation, word choice, spelling, and grammatical mistakes. However, they often overlook global-level errors, ranging from reorganizing the paragraph structures to sharpening the focus. These tools are not designed to provide personalized feedback tailored to individual student needs. Additionally, they mainly focus solely on the errors present in the current task, neglecting to offer Feed-up (suggestions for similar tasks) and Feed-forward (guidance for future tasks). This project aims to address both local-level and global-level errors. The AI assistant is equipped with instructions for providing global-level feedback with Feed-up and Feed-forward. Furthermore, the audio-feedback system includes guidelines for human teachers to offer personalized feedback. It favours teachers in commenting on both the strengths and areas for improvement based on the learners' needs and proficiency levels.
The designs of the AI feedback bot and audio-feedback are suitable for school teachers to incorporate into Process Learning. The teachers can introduce them to their students when they are doing post-writing activities. The ideas and prototypes of this project could highly enhance teaching and learning effectiveness. The effectiveness is reflected on giving more quality feedback on formative writing assignments and boosting students’ motivation on learning writing skills.
To effectively provide detailed feedback on written texts, it is essential for the AI bot to be scalable in order to meet the growing user demand. First, reducing operational costs is crucial. Initially, I used GPT-3.5-Turbo as the base model, providing detailed prompts and supporting documents. However, this approach was expensive, costing nearly 5,000 points on Poe to initiate a conversation with the AI bot. After making several modifications, I switched to Claude-3.5-Sonnet, which is also well-trained in text writing and translation. This change significantly lowered the cost to just over 360 points per message. It could make the AI bot much more accessible for Poe users, who receive 3,000 points every 24 hours.
Second, implementing queuing systems is important. Since the AI bot operates on Poe, we plan to use message queues to manage incoming requests, requiring users to wait when necessary. This will help prevent overload during peak usage times.
To address potential bottlenecks, we will implement database optimization. Poe will utilize systems to optimize database queries and caching to improve data retrieval speed and reduce latency. Additionally, the documents provided to the AI bot will be periodically updated. This ensures that the AI can learn new patterns and types of feedback, helping to avoid repetitive responses.
As a prospective English teacher, I believe we must take the responsibility for helping to achieve the UNESCO Sustainable Development Goal of “Quality Education.” In Hong Kong, teachers in English Language Teaching (ELT) classrooms often fail to provide individualized comments or feedback on students' writing assignments. Many teachers also struggle to reference specific sections of students' handwritten work when providing feedback. This highlights the need for improvement through the introduction of innovative designs and technology.
Effective individualized feedback is crucial for helping students learn how to improve. Comprehensive feedback should consist of three key components: Feed-up (suggestions on similar tasks), feedback (corrections on current tasks), and Feed-forward (directions for future tasks). This approach can significantly enhance students’ performance. The study aims to improve the quality of feedback provided to students, thereby enhancing their writing skills. It is particularly beneficial for lower-ability students by promoting learning equality and equity. These students can gain from receiving more detailed feedback and demonstrations with explanations and examples from an AI assistant. This allows teachers tailor the feedback to cater to their learning needs and deliver detailed explanations through audio feedback. Researches show that audio feedback can help teachers provide more specific and individualized comments. Similarly, the AI feedback assistant and audio-feedback system can further develop the writing skills of high-ability students, pushing them toward excellence.
Overall, this initiative aims to enhance the quality of teachers’ feedback on students’ formative writing assignments by introducing audio feedback and AI writing feedback. The integration of the AI writing feedback assistant and the audio-feedback system on Padlet is driven by the goal of improving teaching and learning effectiveness.
Personal Information Collection Statement (PICS):
1. The personal data collected in this form will be used for activity-organizing, record keeping and reporting only. The collected personal data will be purged within 6 years after the event.
2. Please note that it is obligatory to provide the personal data required.
3. Your personal data collected will be kept by the LTTC and will not be transferred to outside parties.
4. You have the right to request access to and correction of information held by us about you. If you wish to access or correct your personal data, please contact our staff at lttc@eduhk.hk.
5. The University’s Privacy Policy Statement can be access at https://www.eduhk.hk/en/privacy-policy.
- I have read and agree to the competition rules and privacy policy.