Integrating ERNIE Bot in Teaching and Learning Scenarios: Optimizing the Impact of Effective Prompts

Abstract

The ERNIE Bot, a user-friendly and readily accessible AI tool, has the potential to assist teachers in language teaching and learning in various ways. It can catalyze to streamline teachers’ workflow, catering to students’ individual needs and abilities, and enriching the learning experience through a variety of pedagogical approaches. As a large language model (LLM), the effectiveness of prompts crafted by teachers has substantial impacts on the generated outcomes. It is important to acknowledge that certain prompts may not yield the desired results initially, and some may exhibit slightly varied outputs. In such cases, the consideration of regenerating the output or reframing the prompt becomes essential to guide the AI in producing responses aligned with teachers’ expectations. This article, therefore, will explore effective strategies for steering conversations and writing prompts to optimize the ERNIE Bot’s success in creating high-quality teaching and learning materials.
 
Keywords: ERNIE Bot, large language model (LLM), effective prompt(s), regenerate and reframe. 
 

Introduction

The emergence of ChatGPT, or generative AI, has triggered considerable debate within the educational community. While some require students not to use generative AI tools to finish assignments, expressing concerns about potential academic dishonesty and the unreliability of AI-generated responses, others embrace these technologies for their capacity to enhance individuals’ productivity (Mollick & Mollick, 2022; Mollick & Mollick, 2023a). Rather than restricting the use of AI products in education, educators and teachers should actively explore and adopt best practices to maximize the potential of new AI-powered technologies and capitalize on their benefits (Chang et al., 2023).

The ERNIE Bot, developed by Baidu, stands as a ChatGPT-equivalent large language model (LLM) accessible to the general public in China. Despite being a relatively new conversational agent, it is continuously evolving, aiming not only to match but also to surpass ChatGPT in certain aspects. This versatile and intelligent tool, particularly for its language prowess, can significantly enhance the efforts of educators and teachers when utilized wisely and properly. Effectively integrating ERNIE Bot into educational practices can simplify complex tasks, save time, and boost teachers’ work efficiency and productivity. Through leveraging this technology, educators can update their teaching methods, orchestrating more dynamic, engaging, and effective learning activities for their students. 
 

The Importance of Writing Effective Prompts

Prompts are text-based inputs that instruct the LLM to generate specific outputs or execute tasks (Eager & Brunton, 2023; Mollick & Mollick, 2023b). The clarity and precision of the prompt would directly influence the accuracy and specificity of the ensuing output. Hence, teachers’ skill in formulating successful prompts, which takes time to develop and trial, is essential for producing high-quality responses from AI-powered chatbots.

There are several pragmatic strategies for teachers to write effective prompts. Firstly, clear and concise prompts can provide precise instructions to the AI, maximizing their utility and yielding responses aligned with teachers’ expectations. Secondly, well-designed prompts should be relevant to specific learning objectives, catering to appropriate learning levels and potentially motivating and engaging students in the learning process. Lastly, adding constraints or providing contextual information when composing prompts can better accommodate the diverse needs and learning styles of students. By employing succinct sentences, simplified expressions, and precise commands, teachers can create successful prompts that are easy to understand and effective in eliciting favorable replies.

Teachers may also experiment with multiple LLMs, exploring various use cases and interaction techniques to enhance the effectiveness of AI tools. If a prompt does not yield the intended outcome initially, or the AI encounters a processing failure, teachers may reattempt it. Given the inherent unpredictability of AI outputs, repeating the same prompt multiple times may yield different results, and some prompts may not work at all. If a prompt fails, refreshing the conversation and trying again could be a viable solution. If the problem remains, reformulating the prompt or trying a different LLM may prove beneficial. 
 

Scenarios of integrating ERNIE Bot

Scenario 1: Making Crossword Puzzle wordlists 
The crossword puzzle is an activity that teachers may deploy to expand students’ vocabulary range, improve spelling skills, and strengthen social bonds. However, constructing wordlists manually can be time-consuming and laborious. Fortunately, the use of AI tools has significantly improved work efficiency, liberating teachers from the tedious work of manually inputting parts of speech and copying and pasting definitions from online dictionaries. ERNIE Bot as a valuable assistant in automating and streamlining these tasks, provides teachers with additional time to dedicate to their students. Below is a sample prompt, a screenshot of the ERNIE Bot output, and a crossword puzzle handout to demonstrate how this can be done effectively.

Prompt: Here is a list of words: [insert the vocabulary]. First, show the part of speech of each word in the bracket (); and then give definitions of the words for students at CEFR [insert students’ language proficiency] level.
 
 
The preceding example is a quick test with ERNIE Bot, assessing its ability to comprehend instruction and produce desired outcomes. The query presumed that ERNIE Bot would understand the meaning of “CEFR B1”, a parameter intended for generating answers suitable for students at certain language levels. It is important to recognize that the responses from ERNIE Bot may not immediately align with the requirements. Consequently, it is advisable not to accept the initial answer without further refinement, exploration, or probing until a satisfactory response is obtained.
 
Scenario 2: Providing quality examples and exemplars
ERNIE Bot can be applied as a valuable resource supporting teachers to improve students’ writing skills. It furnishes students with high-quality examples and exemplars that demonstrate effective writing techniques and strategies. This scenario examines the feasibility of ERNIE Bot in generating distinctive and meaningful pedagogical materials. In the initial attempt, ERNIE Bot presented a sample answer to a creative writing task by incorporating the required words from the wordlists within seconds. However, it was observed that 14 out of 16 words (underlined in orange) were included in the output, while 2 (circled in green) were omitted, as indicated in the following screenshot. This discrepancy underscores the limitations in ERNIE Bot’s comprehension, highlighting the necessity for further refinement and adjustment to ensure the elicitation of accurate and relevant responses.
 
Prompt 1.0: Use the following words to make a story for students at CEFR [insert students’ language proficiency] level: [insert the vocabulary].
 
 
Since the output of prompt 1.0 did not meet the expectations, the options are either to regenerate it or introduce additional instructions to refine the prompt iteratively. It is worth noting that reframing and enhancing prompts during the conversation can significantly improve the AI output. To test its effectiveness, few constraints are added in prompt 2.0. As depicted in the following screenshot, all 16 words were successfully integrated into the generated response, and the word count amounted to 155, aligning within the prescribed word limit. Therefore, clear and direct instructional texts are crucial for generating quality AI outputs.
 
Prompt 2.0: Use all of the following words to make a story for students at CEFR [insert students’ language proficiency] level: [insert the vocabulary]. Write no more than 200 words.
 
 
Scenario 3: Creating reading comprehension question(s)
ERNIE Bot is instrumental in scaffolding students’ learning and swiftly assessing their reading comprehension. As a starting point, I engaged in a dialogue with the chatbot to fine-tune the specifications and strive for the desired output. If needed, a follow-up prompt may be introduced or further clarification may be requested from ERNIE Bot to elaborate on the response. This iterative process is crucial for delving deeper into the content, allowing the consideration of nuances and perspectives that might have been overlooked. The key lies in continuously testing the requirements fed to the chatbot and tweaking the prompts until they are considered to fit the students’ needs and classroom objectives, thereby elevating the overall learning experience.
 
Prompt: Create [the type of questions] based on the following article: [insert the content].   
          or Create [insert the number] multiple choice question(s) with four options based on the following article: [insert the content].  
 
Follow-up prompt(s): Which is/are the correct answer(s) for the above question(s)? and why?
                                or Why the others are not correct answers?
 
 

Suggestion

ERNIE Bot, similar to any AI chatbot, possesses substantial capabilities, but cannot guarantee alignment of its output with teachers’ educational objectives and expectations. Although it aids teachers in terms of time-saving on specific tasks and provides a framework that can be customized to their needs, it is essential for teachers to meticulously evaluate the content of texts and activities for factual accuracy, appropriateness for learners’ level, and task validity. Teachers should be skeptical of the information provided by these tools, actively supervise the AI’s output, verify it with credible and authoritative sources, think critically, and complement any AI output with their expertise and insights, as AI output may contain inaccuracies, hallucinations, or biases (Mollick & Mollick, 2023a). It is recommended for teachers to build their prompt repository, preserving both the original prompts and the resulting outputs for future reference. 
 

Conclusion

The presence of AI has transitioned from a distant concept to an integral component that permeates various aspects of our daily lives. As educators, it is imperative that we embrace AI as a transformative educational development and unlock new possibilities for teaching and learning. 

AI tools hold the potential to revolutionize education by optimizing teachers’ work efficiency and productivity, improving students’ learning experience, and providing more diverse and tailored learning opportunities. They offer a means to elevate the quality of teaching by simplifying complex topics and lessening time-consuming tasks. Within seconds, teachers can access a selection of optimal responses, ultimately sharpening their skills while utilizing AI as a supplementary tool rather than a substitute. Teachers have the opportunities to experiment, collaborate, and explore the full potential of AI tools by crafting effective prompts, extending beyond conventional practices to foster innovation within the classroom. 
 

Reference

Chang, D.H., Lin, M.P.-C., Hajian, S., & Wang, Q.Q. (2023). Educational design principles of using AI chatbot that supports self-regulated learning in education: Goal setting, feedback, and personalization. Sustainability, 15, 12921. https://doi.org/10.3390/su151712921
Eager,B., & Brunton, R. (2023). Prompting higher education towards AI-augmented teaching and learning practice. Journal of University Teaching & Learning Practice, 20(5). https://doi.org/10.53761/1.20.5.02
Mollick, E.R., & Mollick, L. (2022). New modes of learning enabled by AI chatbots: Three methods and assignments. SSRN. https://ssrn.com/abstract=4300783
Mollick, E.R., & Mollick, L. (2023a). Using AI to implement effective teaching strategies in classrooms: Five strategies, including prompts. SSRN. https://ssrn.com/abstract=4391243
Mollick, E.R., & Mollick, L. (2023b). Assigning AI: Seven approaches for students with prompt. SSRN.  https://arxiv.org/ftp/arxiv/papers/2306/2306.10052.pdf

AUTHOR
Chaonan XU
English Language Centre
School of Languages
XJTLU

DATE
27 February 2024

Related Articles