Generative AI Prompting Techniques
Welcome to our guide on prompting techniques for generative AI chatbots! This page will provide you with practical strategies and examples to help you experiment with the art of AI prompting. Whether you’re planning lessons, designing courses, developing content, or generating images, these techniques will support your efforts.
Content prepared by Jesslyn Wilkinson, Ed Tech Consultant. Writing of this content was aided by Microsoft Copilot, which was used to ideate and draft examples.
On this page
A Prompting Framework
Lo (2023) has conceptualized the CLEAR framework to help you construct prompts.
| Acronym | Description | Example |
|---|---|---|
| Concise | Keep prompts short and to the point. Remove unnecessary words so the AI can focus on the main idea. | “Explain entropy and its significance” instead of “Please provide a detailed explanation about entropy and its significance.” |
| Logical | Organize prompts in a clear, logical order to ensure a smooth flow of ideas. | “Take me through the steps of building a good prompt, starting with how to develop a topic or question to retrieving the prompt result.” |
| Explicit | Be specific about what you want in the output, including format, content, or scope | “Provide a 25-minute podcast script that summarizes entropy, its discovery, and impact.” |
| Adaptive | Be flexible and ready to refine your prompts based on initial results. | If “Provide a 25-minute podcast script that summarizes entropy, its discovery, and impact” is too vague, refine it to “How does the concept of entropy apply in biology and apoptosis?” |
| Reflective | Continuously evaluate and improve your prompts to get the best results. | After receiving the output, assess its quality and make necessary adjustments to your prompt for better results next time. |
Prompting Techniques
Chain-of-Thought:
Also referred to as “conversations,” this is the conversational back-and-forth through which you engage with the chatbot. Because the interaction is a chat log, it can be helpful to progress the chatbot from broader knowledge to more specific. This will also allow you to identify and redirect any hallucinations or inaccuracies.
Example:
(Prompt 1) In this conversation, we’ll be using the BOPPPS lesson planning model, as used by the ISW Network. Describe the model and explain each step in the acronym.
(Prompt 2) Next, use the BOPPPS lesson planning model to draft a [LENGTH OF CLASS] lesson plan on [TOPIC]. The outcome for this lesson is [COURSE OR UNIT OUTCOME]. The participatory activities should have both small group and individual activities. Make effective choices about the activities, so that the lesson plan could be directly implemented by an instructor. The lesson plan should be sectioned into two 50 minute blocks of learning with a 10-minute break in the middle of the lesson.
Persona Patterns:
Chatbots love to adhere to known roles. Leverage this by giving the chatbot an identity to role play as. Suggest a role, job, or identity for the chatbot to take on.
Example:
You’re a college faculty in Ontario, Canada, and an expert in [INDUSTRY] teaching a course on [TOPIC].
Audience:
Give the chatbot a sense of who the intended audience is. Suggest a reading level, language or cultural background, or insight about the audience that would help inform its output.
Example:
The students in this course are primarily/a blend of [CHARACTERISTICS OF LEARNERS] in Ontario, Canada. Create relevant examples for [ACTIVITY] that might be relevant to different students and help build intercultural knowledge.
Zero-Shot:
When you want the chatbot to be more creative, you might use a zero-shot approach. This uses open-ended prompts and is less specific about the output. The prompt should still be specific in terms of topic and objective.
Example:
You’re teaching a lesson on [TOPIC]. The outcome of this lesson is [COURSE OR UNIT OUTCOME]. An effective bridge-in for a lesson should take 5 minutes of class time or less. List 5 different ideas for bridge-in activities for this lesson.
Few-Shot:
When you want the chatbot to follow a structure or example, you might use a few-shot approach. This uses examples to structure the chatbot’s outputs.
Example:
Let’s develop some multiple-choice questions. Use the following format to develop the questions:
What is the name of a mammal that is large, grey and has a long trunk?
- A dinosaur
- An elephant
- A rhino
- A whale
Correct answer: B. An elephant
Using the example above as a template, create [# OF QUESTIONS] [ASSESSMENT/PRE_TEST/POST-TEST] questions on [TOPIC or OUTCOME].
Prompt Examples
In this section, we’ll share examples of prompts that could help you achieve some target teaching-related tasks with generative AI chatbots. If you have additional examples you’d like to see here, please contact us.
Email Writing
Scenario: You would like help crafting email replies to students.
Initial Prompt: “Compose an email to students reminding them of the upcoming deadline for [ASSIGNMENT NAME]. Include the due date, submission instructions, and any resources available to help them complete the assignment. Use a professional and friendly tone.”
Extension: “Add a section encouraging students to reach out if they have any questions or need further assistance.”
Generating Outcomes
Scenario: You are developing learning outcomes for a new course.
Initial Prompt: “Generate a list of learning outcomes for a course on [COURSE NAME]. The course description is [DESCRIPTION]. Follow Canadian standards and frameworks such as [INDUSTRY STANDARDS, ORGANIZATIONS OR FRAMEWORKS]”
Extension: “Align each learning outcome with a corresponding assessment method to measure student achievement.”
Lesson Planning
Scenario: You want to create a detailed lesson plan for a new unit in your course.
Initial Prompt: Apply the BOPPPS lesson planning model to create a detailed lesson plan for [TOPIC]. The target outcome for this lesson is [OUTCOME]. The lesson plan should last [# MINUTES].
Prompt Extensions: “Next, generate a list of 5 bridge-in activities that align with the outcome.” OR “In the past, my students have enjoyed these types of activities: [ADD EXAMPLES]. Develop one small group and one independent activity that aligns with these types of activities and the target outcome. Ensure the activities align with active learning principles.”
Course Design
Scenario: You’re designing a new course and need a comprehensive content outline.
Initial Prompt: “You are a subject matter expert on the topic of [TOPIC]. You’re developing a 15-week course on [COURSE NAME]. The course description is [COURSE DESCRIPTION]. It aims to support [OUTCOMES]. Generate an outline of the relevant content for this course, including weekly topics and potential assignments. Week 8 of the semester should be a break week, with no assessments or learning tasks assigned.”
Prompt Extensions:
- “Build on Week 1 [TOPIC]. The target outcome is [COURSE AND/OR UNIT OUTCOME(S)]. Create a plan for this week’s learning.”
- “Reconsider the proposed assessments. Develop an assessment that targets [COURSE AND/OR UNIT OUTCOME(S)]. The assessment should/should not incorporate student use of generative AI as a collaborator.”
Multiple Choice Questions
Scenario: You want to create multiple-choice questions for a quiz.
Initial Prompt: “Generate 20 multiple-choice questions for a quiz on [OUTCOME(S)]. Each question should include three incorrect but plausible options and one correct option. Use the following template for the questions:
What is the name of a mammal that is large, grey and has a long trunk?
- A dinosaur
- A whale
- An elephant
- A rhino
Correct answer: An elephant
Prompt Extensions:
- “These questions do not yet address [OUTCOME]. Generate 10 additional questions aligned to this outcome. Continue to use the question template.”
Video Script
Scenario: You need to develop engaging video content for your course.
Initial Prompt: “Generate a script for a 7-minute video on [TOPIC]. Recommend visuals to include on slides to complement key moments in the script.”
Extension:
- “Add real-world applications or examples.”
- “Reduce the script length by 10%. Ensure you do not eliminate the key concepts of the video.”
Image Generation
When generating images, it’s recommended to be additionally cautious, as image generators can unintentionally reproduce copyrighted material. Be cautious when using Copilot for this purpose, and if the intention is to openly license the images, it may be beneficial to explore using Adobe Firefly instead, provided by the college through our Adobe licensing agreement. Your manager must approve your use of Adobe Creative Cloud. Request access to Adobe via IT’s online form.
Scenario: You want to create visuals to use in a presentation, OER, or other course development process. In this case, you’ll want to match the output you’re looking for to how you prompt the chatbot.
It’s worth noting that it is best to be very cautious when generating realistic images of humans. Most chatbots do not reflect the diversity and variety of the human experience in their outputs. In many cases, you may need to directly prompt for more diverse images or particular ethnic, gender, or cultural respresentations.
Initial Prompt: “4K photorealistic image of [PEOPLE] doing [ACTIVITY].”
Extension: “Adapt this image to include a more diverse representation of Canadian society.”
Student Engagement
Scenario: You aim to increase student engagement through interactive activities.
Initial Prompt: “Suggest 5 active learning activities for a lesson on [TOPIC]. The activities should take approximately [# MINUTES]”
Extension: “Refine [IDEA FROM OUTPUT] to incorporate more peer-to-peer interaction.”
Assessment Design
Scenario: You need to create effective assessment tools for your course.
Initial Prompt:
- “Design an assessment to evaluate [OUTCOME(S)]. The assessment should include [TYPES OF QUESTIONS OR TASKS]. Ensure that there is a range and variety of types of questions asked. The questions should require connection to real-world examples or applications.”
- “Design a real-world, authentic assessment to match [OUTCOMES]. The assessment could support multiple means of representation and UDL principles. The amount of time and effort required should match the overall grading allotment of [% OF TOTAL GRADE].”
Rubric Criteria
Scenario: You need to create a detailed rubric for assessing a research paper.
Initial Prompt: “Develop a detailed rubric for assessing a research paper on [TOPIC]. The rubric should include criteria such as clarity of thesis, depth of research, organization, grammar and style, and adherence to formatting guidelines. Each criterion should have a scale from 1 to 5, with descriptions for each level.”
Extension: “Include examples of what constitutes a level 5 and a level 1 for each criterion.”
References and Further Reading
By using these prompting techniques and strategies, you can effectively harness the power of generative AI to enhance your teaching and course development. Experiment with different prompts and refinements to find what works best for your specific needs. The resources below can further complement your learning and practice.
AI for Education. (2023). Prompt Library.
Coursera. (2023). Prompt Engineering for Educators. (Free Course)
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.
Lo, Leo, S. (2023). The CLEAR path: A Framework for Enhancing Information Literacy through Prompt Engineering. The Journal of Academic Librarianship, 49(4).
LinkedIn Learning. (2022). Introduction to Prompt Engineering for Generative AI. Accessed via Conestoga’s Library Services, July 2024.
Mollick, E. (2023). Working with AI: Two Paths to Prompting. One Useful Thing. Retrieved July, 2024.
UC Davis Library. (2023). Prompt Engineering. Generative AI for Teaching, Research and Learning.