Skip to content

Newsletter#22: Teaching with AI: Optimizing Responses through Prompt Engineering

截屏2025-03-27 19.39.43

Teaching with AI: Optimizing Responses through Prompt Engineering

We are excited to introduce our new bi-weekly newsletter series, "Teaching with AI," designed to explore the integration of artificial intelligence in educational practices. This series complements our Canvas course, "Application of AI in Teaching," providing practical insights and strategies for educators. Our inaugural issue focuses on Prompt Engineering, a crucial skill for effectively utilizing generative AI tools in the classroom.

Introduction to Prompt Engineering

Firstly, what is a prompt?

A prompt is a command sent to a large model, such as "Tell a joke", "Write a Snake game in Python", or "Compose a poem".

Why do we need to learn prompts?

Large models accept only one type of input: prompts.

What is Prompt Engineering?

  • Prompt engineering is the process of creating prompts, asking questions, or guiding language models like ChatGPT to generate outputs.

Prompt engineering has a low entry barrier but a high ceiling, which is why some people jokingly refer to prompts as "spells".

1280X1280-2      1280X1280 (1)-1  ChatGPT 4o（Left）  DeepSeek（Right）

What are the differences between a prompt, prompt engineering, and prompt engineers?

  • Prompt is the "programming language" of the AGI era.

  • Prompt engineering" is the "programming" of the AGI era.

  • Prompt engineers" are the "programmers" of the AGI era.

 

Why is learning prompt engineering important for us in the AGI era?

In the AGI era, mastering prompt engineering is as essential as using a mouse and keyboard. This skill unlocks AI’s full potential, enhancing efficiency, creativity, and problem-solving in the digital age.

3-3

 
截屏2025-03-27 15.10.46 

   

In this newsletter, we present 10 official ChatGPT prompting techniques based on OpenAI’s Strategies for Getting Better Results, complete with examples. These techniques have been compiled to help users get better responses from ChatGPT. They can also serve as a reference when using other Generative AI tools.

10 ChatGPT Prompting Techniques Based on OpenAI's "Strategies for Getting Better Results"

ChatGPT Prompting Technique 1: Role-Playing

  • When asking a question, take on the role of the user or the question proposer. This helps frame more specific and clear inquiries, leading to more precise and targeted responses.

  • First, define the role—this essentially narrows down the problem domain at the beginning, reducing ambiguity.

  • Example:

4-Mar-27-2025-07-13-47-4975-AM

 

ChatGPT Questioning Techniques 2: Providing Contextual Information

  • Include relevant background information about the topic being discussed. This helps the AI understand the specific meaning of the question, leading to more accurate answers.

  • A good way to provide context is to make your question specific enough by including details about what you are trying to do, the issue you are encountering, and the kind of result you hope to achieve.

  • Example:

5-2

  • Incorrect example:

    • How good is this movie?

 

ChatGPT Questioning Techniques 3: Providing Keywords

Keywords are the most important words or phrases that describe your question or topic. They help AI better understand your needs and provide more precise answers.

Examples:

  • I am learning Python and encountered an error called 'Indentation Error.' What should I do?

  • Cell F2 contains an 18-digit ID number. Based on the 17th digit, determine gender: if the 17th digit is odd, return 'Male'; otherwise, return 'Female.' Write an Excel formula.

 

ChatGPT Questioning Techniques 4: Providing Prior Information

  • Providing prior information ensures that your question is more specific, clear, and includes enough context for ChatGPT to better understand your intent.

  • Structure: A piece of content + a question based on that content.

  • Example:

    • I am studying calculus, but I am having difficulty calculating definite integrals. Specifically, I am trying to compute the definite integral of a function over a given interval:

6-Mar-26-2025-07-07-25-6466-AM

Can you guide me through the process?

  • Incorrect Examples:

    • "How do you calculate a definite integral?"

    • "Can you explain the concept of a definite integral?"

 

ChatGPT Questioning Techniques 5: Limiting Length

  • Forcing ChatGPT to be concise ensures that it conveys richer information within a limited word count or number of paragraphs.

  • Examples:

    • I will input a long passage. Do not explain, discuss, or respond until I ask a question. Understood?

    • "Based on the content above, write a summary with the topic sentence '2023 Annual Work Summary.'

      • Focus on my achievements

      • Follow this structure: Work Overview, Tasks, Progress, Achievements, Completion Status, Performance Evaluation, Lessons Learned, Outlook & Recommendations, Next Steps, Closing Remarks

      • Must be at least 1000 words.

    • Carefully read the document and write a summary of it in 300 words.

 

ChatGPT Questioning Techniques 6: Providing Examples

  • Providing examples when necessary, a technique known in academia as Few-Shot Learning or In-Context Learning, which significantly improves output accuracy.This technique involves mimicking a given example, allowing ChatGPT to generate responses that align with a specific writing style or structure.

  • Examples:

    • Read the above copy and analyze its structure, language features, and emotional tone

    • Based on the writing characteristics of the above copy, rewrite a promotional text about postgraduate equivalency programs. Add emojis.

    • Carefully read the document and analyze its writing structure, style, language features, and emotional tone.

7-2

 

ChatGPT Questioning Techniques 7: Controlling Output Style

  • Describe the desired style and format of the output, guide the model to return only the required information, and structure the output to facilitate automatic parsing by subsequent modules.

  • Examples:

    • Adjusting tone: “Rewrite the response in a more formal/professional/casual tone.”

    • Controlling length and depth: “Provide a brief explanation in 100 words” / “Give a detailed analysis in 500 words.”

    • Specifying emotional color: “Write the response with an encouraging/neutral/sympathetic tone.”

    • Adjusting language style: “Rewrite this in the style of classical Chinese prose.”

    • Formatting output:

      • Excel format: “List the advantages and disadvantages of AI in an Excel table.”

      • Markdown format: “Generate the response in Markdown format.”

 

ChatGPT Questioning Techniques 8: Let Me Think

  • This technique encourages deeper reflection, prompting ChatGPT to pause briefly before responding. This can lead to more thoughtful and comprehensive answers.

  • Example:

8-Mar-26-2025-07-11-10-6709-AM

 

ChatGPT Questioning Techniques 9: Specify the Target Audience

  • This means clearly identifying the intended audience or specific user needs in your question so that ChatGPT can provide more targeted and personalized responses.

    • Specify a particular group

    • Identify specific needs

    • Adapt to specific scenarios

  • Examples:

    • Explain the mechanism of large language models to a programmer with ten years of work experience.

    • Explain the mechanism of large language models to a third-year undergraduate student with weak math skills.

  • Incorrect Example:

    • Explain the mechanism of large language models

 

ChatGPT Prompting Tip #10: Review Your Question

  • Ensure accuracy, relevance, and reliability. This step is especially important because large language models like ChatGPT may occasionally produce hallucinations—confident-sounding but incorrect or fabricated information.

To further understand the phenomenon of hallucinations, here are some pertinent studies and reports:

1-1-2

1-2-1

1-3
1-4
 
 

Explore the Canvas Course: 'Application of AI in Teaching'

This comprehensive course offers a wealth of resources, including pertinent literature, online video tutorials, interactive discussions, assessments, and assignments. Upon successful completion, participants will receive a badge from the Center for Teaching and Learning (CTL), recognizing their proficiency in applying AI in educational settings.

The course is scheduled to launch next fall. We warmly invite faculty members to share their experiences and insights regarding the practical application of AI in teaching. Your contributions will greatly enrich our collective understanding and foster a collaborative learning environment.

Reference:

OpenAI. (n.d.). Prompt engineering. OpenAI. Retrieved March 27, 2025, from https://platform.openai.com/docs/guides/prompt-engineering

 

Author:  Yirui (Sandy) Jiang