See discussions, stats, and author profiles for this publication at: [Link]

net/publication/385073803

Engineering ChatGPT Prompts for EFL Writing Classes

Article in International Journal of TESOL Studies · November 2023

DOI: 10.58304/ijts.20230405

CITATIONS READS

16 232

1 author:

Jerry Yung Teh Huang

Kansai University
8 PUBLICATIONS 34 CITATIONS

SEE PROFILE

All content following this page was uploaded by Jerry Yung Teh Huang on 19 October 2024.

The user has requested enhancement of the downloaded file.

International Journal of TESOL Studies (2023)
Vol. 5 (4) 73-79 [Link]

Article

Engineering ChatGPT Prompts for EFL Writing Classes

Jerry Huang
Kansai University, Osaka, Japan

Received: 28 August 2023; Accepted: 8 September 2023; Published: 5 November 2023

Abstract
In the realm of EFL education, the labor-intensive task of providing feedback on student writing
prompts educators to seek alternatives. Peer feedback and computer-based systems, such as
ChatGPT, may offer solutions. ChatGPT’s AI-driven feedback relies on Large Language Models, offering
real-time assessments to alleviate the workload of teachers. Crafting precise prompts is pivotal to
harnessing ChatGPT’s potential, highlighting the necessity of cultivating digital literacy in the AI-
driven educational age. Instead of replacing traditional teaching, AI tools like ChatGPT can become
collaborators, enhancing personalized learning experiences for educators and students alike. This
article will demonstrate effective prompts for writing classes.

Keywords
ChatGPT, Generative-AI, Prompts, EFL writing, feedback

1 Introduction

Teaching English as a Foreign Language (EFL) writing can be an enormous burden on educators,
often resulting in an overwhelming workload (Chen, 2021). The task of providing individualized and
constructive feedback to a classroom full of diverse learners is both time-consuming and mentally taxing.
However, in light of these challenges, alternative methods such as peer feedback and computer-based
feedback have emerged as viable options (Alnasser, 2018). Utilizing peer feedback motivates students
to actively engage in the learning journey while providing them with valuable insights from their fellow
classmates (Zhang et al, 2020). Similarly, computer-based feedback systems, equipped with advanced
algorithms, can offer instant evaluations and tailored suggestions, alleviating some of the teacher’s grading
responsibilities, as demonstrated by Mizumoto and Eguchi (2023), highlighting the feasibility of automated
essay scoring. These alternatives not only ease the teacher’s workload but also promote collaborative
learning and independence among students, making them valuable tools in the EFL writing classroom.
ChatGPT can serve as an excellent computer-based feedback tool for EFL writing classes. Its
advanced language capabilities and natural language understanding make it adept at providing precise
and personalized feedback to students. ChatGPT can offer instant evaluations of grammar, vocabulary,
and sentence structure, helping learners identify and correct their mistakes in real-time. Additionally,
it can suggest alternative phrasing and expressions, enriching students’ language skills (Kohnke et al.,

Email: jerryyhuang@[Link]
74 International Journal of TESOL Studies 5 (4)

2023). Furthermore, ChatGPT allows students to access feedback whenever they need it, promoting self-
directed learning and improving overall writing skills (Dai et al., 2023). As an AI-powered feedback
resource, ChatGPT not only lightens the teacher’s workload but also enhances the quality of guidance
provided to EFL students, contributing to more effective and efficient language learning experiences.
ChatGPT, an advanced AI language model from OpenAI, is founded on the GPT-3.5 architecture.
GPT stands for Generative Pre-trained Transformer, is a type of AI model that can generate text and is
built upon the Transformer architecture. It excels in facilitating natural conversations, understanding user
inputs, and generating contextually fitting responses. With extensive training on diverse text sources,
ChatGPT possesses a broad knowledge base. Its neural network framework comprehends queries and
nuances, crafting human-like replies, making it versatile for support, content creation, and knowledge
sharing. Despite occasional limitations and inaccuracies, ChatGPT exemplifies AI’s evolution in
understanding and generating human language, marking a new era of interactive technology.
The interface of ChatGPT is user-friendly and straightforward. On the main page, users encounter a
text box to input prompts or questions, accompanied by a “send” button to trigger responses. ChatGPT
is accessible for free during periods of low demand, and a paid version (ChatGPT 4.0) is available for
quicker responses and priority access during peak times, alongside new features.
ChatGPT has already demonstrated success in language writing classes. Notably, Dergaa et al. (2023)
conducted a study that highlights ChatGPT’s capacity to enhance academic writing, while Schmidt-Fajlik
(2023) found it to be especially beneficial for Japanese English Language Learners (ELLs), particularly in
the realm of grammar. While valuable for language learning, ChatGPT’s limitations, such as its potential
to miss nuances and promote over-reliance, may hinder students’ creativity, critical thinking, and raise
plagiarism concerns (Marr, 2023; Granados, 2022; Stewart, 2023). It also is important to acknowledge that
while ChatGPT holds promise as a language tool, it may occasionally generate results that lack depth, are
overly generic, and fail to produce informative content (Barrot, 2023; Teng, 2023). As advised by Barrot
(2023), the prudent approach involves using ChatGPT as a supplementary resource rather than relying
on it as the sole feedback source. This balanced approach ensures that with thoughtful guidance from
educators, ChatGPT can effectively contribute to the development of crucial writing skills and practices.
In this article, I will demonstrate how students can effectively utilize provided prompts to acquire
valuable feedback on their writing. During the field testing of these prompts, I opted for the no-cost
version of ChatGPT 3.5. Despite encountering sporadic instances of brief usability interruptions,
typically lasting only a few minutes, such disruptions were relatively infrequent. As previously noted,
ChatGPT 3.5 can be accessed for free during periods of low demand, while a paid version (ChatGPT 4.0)
offers faster responses and priority access during peak times, accompanied by additional features. This
study utilizes ChatGPT 3.5 to investigate an emerging trend in EFL education where readers can readily
replicate the prompts and employ them without incurring any costs, contributing to broader and more
accessible implementation for educational purposes.
RQ1: What prompts can be helpful for effective writing feedback?
RQ2: How should prompts be organized to target different parts of an essay?

2 Prompt: Generic Example

For ChatGPT to offer effective feedback, the selection of an appropriate prompt is crucial. A prompt in
ChatGPT initiates responses and requests information through text input. To achieve this, the prompt
should be as direct as possible. Furthermore, incorporating rubrics into the prompt becomes essential
as it allows educators to streamline their grading expectations (Su et al., 2023). These rubrics serve as
clear guidelines, directing ChatGPT to focus on specific writing aspects like grammar, coherence, or
persuasion. When a student provides a prompt along with a loaded rubric, ChatGPT can precisely tailor
Jerry Huang 75

its responses to meet the specified criteria. This dual approach not only simplifies the feedback process
but also ensures that the evaluation aligns seamlessly with the learning objectives of the assignment.
In the realm of Large Language Models (LLMs), two primary prompting techniques exist: zero-
shot and few-shot prompting. Zero-shot prompting enables the LLM to generate responses or perform
tasks without the need for task-specific training. It relies on its pre-existing knowledge base to provide
meaningful results and is often used specifically for information retrieval. In contrast, few-shot
prompting extends this approach by supplying the model with examples or demonstrations, aiding its
comprehension of the task or context. Consequently, this augmentation leads to more accurate and
contextually relevant information generation. In essence, zero-shot prompting leverages pre-existing
knowledge in LLMs for task execution, while few-shot prompting further enhances performance by
providing specific examples or demonstrations.
To enable ChatGPT to provide feedback effectively, neither zero-shot nor few-shot prompts are
suitable. To clarify, the prompt should explicitly request a specific action. Here is an example of a prompt
that is user-friendly for both students and teachers.

Example Prompt 1
Please review my paragraph, provide feedback on structure, grammatical errors, clarity, coherence, logic, and
use of appropriate transition and cohesive device. I am not seeking rewrite, only feedback.

(Paste Your Paragraph here)

The prompt instruction is straightforward and clear, providing precise guidance on ChatGPT’s intended
action. Students can effectively utilize it to enhance their writing based on the feedback they received.
It is important to note that failing to include phrases like “I am not seeking a rewrite, give me only
feedback” can have unintended consequences. When omitted, ChatGPT not only offers feedback in
accordance with the specified criteria but may also undertake a comprehensive revision of the paragraph.
In such a scenario, students might be tempted to simply copy and paste the resultant text, potentially
bypassing the core objectives of the assignment. This issue highlights one of the major concerns
associated with the utilization of ChatGPT (Today, 2023).

3 Prompt Organization: An Illustrative Example

A specialized technique, known as “chain-of-thought prompting” (Wei et al., 2022), adds another
dimension. This method guides LLMs through complex inference tasks by providing intermediate steps,
which elevates the model’s accuracy when paired with few-shot prompting. The key takeaway is that the
effectiveness of an LLM improves with the inclusion of more specific examples in the prompt.
For this reason, to create an effective prompt for writing evaluation, it is advisable to utilize a prompt
that includes specific examples, which serve as criteria. The prompt should be structured into three
distinct sections: (1) task, (2) criteria, and (3) “my paragraph.” The “task” section serves as the primary
instruction, outlining precisely what ChatGPT is expected to do. It is crucial that this section is detailed
and specific, as merely asking ChatGPT to provide feedback in a general sense can be too vague.
Moreover, incorporating considerations such as feedback tailored to different language proficiency levels,
like CEFR, is beneficial for accommodating various EFL learner levels.
Within the “criteria” section, educators can effectively align their expectations with grading rubrics.
In academic writing, the widely utilized five-paragraph structure serves as a common framework.
However, it is crucial to note that each distinct paragraph should be evaluated differently. Take the
introduction paragraph, for example. In this case, the rubrics should specifically focus on elements like
76 International Journal of TESOL Studies 5 (4)

the hook or attention-grabber, the provision of background information or context, followed by the thesis
statement. Beyond this basic structural assessment, it is imperative to scrutinize aspects of organization,
clarity, and coherence for a comprehensive and dynamic introduction paragraph.
By incorporating the aforementioned task and criteria and then inserting the student’s work under the
designated “my paragraph” section, this approach is exemplified in the sample provided below.

Example Prompt 2
### Task ###
Based on the following criteria, please review the following paragraph and provide specific feedback that will help
improve it. Provide your feedback at the B1 level of the CEFR. Instead of rewriting the paragraph, give specific
examples and guidelines on how to revise. I am not seeking rewrite, give me only feedback.

### Criteria for the Introduction Paragraph###

1. Hook/Attention-Grabber: The introduction engages the reader with a compelling hook or attention-grabbing
statement.
2. Background Information/Context: The introduction provides relevant background information or context to set the
stage for the essay topic.
3. Thesis Statement: The introduction includes a clear and concise thesis statement that presents the main argument
or purpose of the essay.
4. Organization: The introduction demonstrates a logical and coherent structure, guiding the reader smoothly into
the main body of the essay.
5. Clarity and Coherence: The introduction is written in a clear and coherent manner, effectively conveying the
purpose and direction of the essay.

### My paragraph ###

(Paste Your Paragraph here)

Certainly, the previously mentioned prompt may not be as effective when applied to body paragraphs,
given the distinct criteria for evaluating them. In contrast, body paragraphs require a different set
of criteria for comprehensive assessment. Consider the following example criteria, which not only
encompass aspects like the topic sentence, development and support, organization and structure, and
transitions but also delve into higher-order skills such as analysis, critical thinking, and maintaining
clarity and coherence.

Example Prompt 3
### Criteria for the Body Paragraphs###
1. Topic Sentence: Each body paragraph begins with a clear and focused topic sentence that introduces the main
idea of the paragraph.
2. Development and Support: Each body paragraph provides sufficient and relevant evidence, examples, or
arguments to support the main idea.
3. Organization and Structure: Each body paragraph demonstrates a logical and coherent structure, with ideas
presented in a well-organized and connected manner.
4. Transitions: Each body paragraph uses appropriate transitions and cohesive devices to create smooth flow and
connections between sentences and ideas.
5. Analysis and Critical Thinking: Each body paragraph includes thoughtful analysis, interpretation, or evaluation
of the evidence or examples presented.
6. Clarity and Coherence: Each body paragraph is written in a clear and coherent manner, effectively conveying the
ideas and arguments.
Jerry Huang 77

Finally, when examining the concluding paragraphs, attention should be directed towards several key
components. These include the restatement of the thesis, a concise summary of the key points discussed,
a closing thought that leaves a lasting impression, establishing a connection with the introduction, and
culminating with a strong concluding sentence. The provided example criteria below outline the criteria
for evaluating these crucial aspects within concluding paragraphs.

Example Prompt 4
### Criteria for the Concluding Paragraphs###
1. Restatement of Thesis: The concluding paragraph includes a clear restatement of the thesis statement,
reminding the reader of the main argument or purpose of the essay.
2. Summary of Key Points: The concluding paragraph summarizes the key points or main ideas discussed in
the essay’s body paragraphs.
3. Closing Thought: The concluding paragraph provides a thoughtful closing thought or final statement that
leaves a lasting impression on the reader.
4. Connection to the Introduction: The concluding paragraph effectively connects back to the introduction,
bringing the essay full circle and providing a sense of closure.
5. Concluding Sentence: The concluding paragraph ends with a strong concluding sentence that wraps up
the essay and leaves the reader with a sense of completion.

The crafted prompt showcased earlier demonstrates its versatility and adaptability. By offering a clear
instruction followed by well-defined criteria, ChatGPT efficiently handles the feedback process. Notably,
these criteria can be adjusted to align with the goals of different types of writing tasks. Undoubtedly,
creating prompts tailored to the specific needs of a writing course can be a time-intensive process.
Nevertheless, this effort is undeniably valuable, as it lays the groundwork for effective AI-driven
feedback, playing a pivotal role in enhancing students’ writing skills and overall learning experience.
The prompts discussed here did not emerge overnight. Rather, they were carefully crafted by the
author through a process of trial and error, tailored to meet specific objectives while conducting classes.
Ongoing empirical research is currently examining these prompts, and the validation results will be
reported in due course. Meanwhile, based on my own experience as a practitioner who has tested this
with learners, I believe they have proven to be effective in classroom teaching.

4 Discussion

Using ChatGPT for peer feedback holds the potential to significantly reshape EFL writing classes. For
one thing, ChatGPT’s AI-driven capabilities can provide quick and consistent feedback to students,
fostering a sense of self-efficacy and independence in their learning process (Dai et al., 2023). It can
also help students identify common errors and improve their writing skills. Furthermore, ChatGPT can
accommodate different proficiency levels, tailoring feedback to beginner, intermediate, and advanced
learners, making it adaptable for diverse EFL contexts. ChatGPT can also serve as a valuable writing
assistance, suggesting alternative word choices, sentence structure, and idiomatic expression, enriching
students’ language skills (Kohnke et al., 2023). These advantages undoubtedly benefit students and can
help alleviate the workload for teachers, especially in larger class sizes.
However, it is crucial to acknowledge ChatGPT is not infallible and may sometimes miss nuances,
context, or the emotional aspect of feedback that a human peer or teacher can provide (Marr, 2023).
Over-reliance on AI for peer feedback could potentially stifle creativity and critical thinking in students
(Granados, 2022) and raise concerns about “high-tech plagiarism” (Stewart, 2023) if students use
78 International Journal of TESOL Studies 5 (4)

ChatGPT to generate or heavily edit their writing pieces without proper citation. Therefore, a balanced
approach that combines the benefits of ChatGPT with human oversight and guidance is essential to
harness its potential effectively. In doing so, educators can empower students to become more adept
to discerning and using AI-generated feedback in their learning journey while preserving the essential
human elements of education. This review showcased the creation of prompts as a strategy to mitigate the
disadvantages of ChatGPT utilization in EFL writing classes while still preserving its core advantages.

5 Concluding Remarks

Employing a meticulous prompt in educational settings when utilizing ChatGPT holds paramount
importance (Tseng & Warschauer, 2023). The quality and precision of the prompt serve as a crucial
foundation for the type of responses generated by the AI model. A well-crafted prompt not only ensures that
the desired information or feedback is obtained but also facilitates the generation of relevant and insightful
content. A careful prompt takes into account the specific context, objectives, and desired outcomes of
the educational interaction, leading to more accurate and meaningful AI-driven responses. Moreover, a
thoughtful prompt minimizes the risk of generating irrelevant or misleading outputs, contributing to a more
effective and productive engagement between learners, educators, and the AI technology.
In the era of AI-driven education, both students and teachers stand to gain immense benefits from this
technological revolution. However, at the heart of this transformative shift lies the essential prerequisite
of digital literacy. The ability to navigate, comprehend, and effectively utilize AI tools and platforms
is paramount. When it comes to AI-driven educational tools like ChatGPT, the art of crafting a prompt
takes center stage. It is not just about copying and pasting questions but rather about formulating prompts
that foster genuine learning experiences. In light of this, the saying “If you can’t beat them, join them”
becomes particularly relevant. Embracing AI as a powerful ally in education can open new frontiers for
personalized, efficient, and enriching learning journeys. It is not merely about adopting technology for
the sake of it; it is about harnessing its potential to enhance the educational process for both educators
and students alike.

References

Alnasser, S. M. N. (2018). Exploring student-writers’ views on replacing teacher feedback with peer
feedback and computer-based feedback. Arab World English Journal, 9(3), 345–366. [Link]
org/10.24093/awej/vol9no3.23
Barrot, J. S. (2023). Using ChatGPT for second language writing: Pitfalls and potentials. Assessing
Writing, 57, 100745. [Link]
Chen, J. (2021). Research on the effect of peer feedback training in English writing teaching—A case
study of students in business English major. English Language Teaching, 14(6), 12. [Link]
org/10.5539/elt.v14n6p12
Dai, Y., Liu, A., & Lim, C. P. (2023). Reconceptualizing ChatGPT and generative AI as a student-
driven innovation in higher education. Procedia CIRP, 119, 84–90. [Link]
[Link].2023.05.002
Dergaa, I., Chamari, K., Zmijewski, P., & Saad, H. B. (2023). From human writing to artificial
intelligence generated text: examining the prospects and potential threats of ChatGPT in academic
writing. Biology of Sport, 40(2), 615–622. [Link]
Granados, N. (2022, January 31). Human Borgs: How artificial intelligence can kill creativity and make
Jerry Huang 79

us dumber. Forbes. [Link]

artificial-intelligence-can-kill-creativity-and-make-us-dumber/?sh=320ec11821a2
Kohnke, L., Moorhouse, B. L., & Zou, D. (2023). ChatGPT for language teaching and learning. RELC
Journal. 003368822311628. [Link]
Mizumoto, A., & Eguchi, M. (2023). Exploring the potential of using an AI language model for
automated essay scoring. Research Methods in Applied Linguistics, 2(2), 100050. [Link]
org/10.1016/[Link].2023.100050
Marr, B. (2023, March 3). The top 10 limitations of ChatGPT. Forbes. [Link]
bernardmarr/2023/03/03/the-top-10-limitations-of-chatgpt/?sh=3a52aacc8f35
Schmidt-Fajlik, R. (2023). ChatGPT as a grammar checker for Japanese English language learners:
A comparison with Grammarly and ProWritingAid. AsiaCALL Online Journal, 14(1), 105–119.
[Link]
Stewart, J. (2023). Noam Chomsky says ChatGPT is a form of “High-Tech plagiarism” My Modern Met.
[Link]
high-tech,to%20avoid%20doing%20-the%20work
Su, Y., Lin, Y., & Lai, C. (2023). Collaborating with ChatGPT in argumentative writing classrooms.
Assessing Writing, 57, 100752. [Link]
Tardy, C. M. (2012). Chapter 7. A rhetorical genre theory perspective on L2 writing development. In R.
Manchón (Ed.), L2 writing development: Multiple perspectives (pp. 165–190). DE GRUYTER.
[Link]
Teng, M. F. (2023). Scientific writing, reviewing, and editing for open-access TESOL journals: The role
of ChatGPT. International Journal of TESOL Studies. [Link]
Tseng, W., & Warschauer, M. (2023). AI-writing tools in education: If you can’t beat them, join them.
Journal of China Computer-Assisted Language Learning. [Link]
Wei, J., Wang, X., Schuurmans, D., Bosma, M., Ichter, B., Xia, F., Chi, E., Le, Q., & Zhou, D. (2022).
Chain-of-thought prompting elicits reasoning in large language models. ArXiv [CsCL]. [Link]
org/10.48550/ARXIV.2201.11903
Zhang, J., Kuusisto, E., Nokelainen, P., & Tirri, K. (2020). Peer feedback reflects the mindset and
academic motivation of learners. Frontiers in Psychology. [Link]

Jerry Huang is an English language instructor at Kansai University, originally from Los Angeles. He
holds a master’s degree from Hyogo University of Teacher Education and is currently pursuing a Ph.D.
in foreign language education and research at Kansai University. Jerry’s teaching expertise lies in English
language courses, while his research interests encompass motivation, AI, ChatGPT, and Languages Other
Than English (LOTE).

View publication stats