AI tools like ChatGPT, Google Gemini, and Microsoft CoPilot are changing the landscape of education and content creation. For instructors, these tools offer exciting possibilities for streamlining course development, enhancing learning materials, and engaging students in new ways. The key to unlocking this potential lies in developing the skill of prompt writing.
What is Generative AI and Prompt Writing?
Generative AI is a type of artificial intelligence model that creates various types of content, including text, images, code, media, and more. These models learn from existing datasets and can generate responses based on the instructions or "prompts" that you provide.
AI's Language
AI's ability to understand our language comes from two key technologies: Natural Language Processing (NLP) and Machine Learning (ML). NLP helps AI grasp the meaning behind your words, even if you write casually. ML allows the AI to constantly learn and improve its responses based on the prompts it receives.
Three Basic Strategies
You can achieve a lot with simple prompts, but the quality of results depends on how much information you provide the tool. Effective prompt writing involves crafting clear, specific, and detailed instructions to guide the AI toward your desired outcomes.
Provide Context
AI performs best when it understands the big picture.
If you ask, "Write a summary of Shakespeare's Hamlet." AI will give a basic answer.
If you add context like, "Write a summary of Hamlet for ESL students, focusing on key plot points and using simplified language." AI will tailor its response accordingly.
Be Specific
The more details you provide, the better the AI tool performs. Think of it as giving AI a clear roadmap to your desired output.
Instead of "Write a lesson plan on climate change," you could try "Write a 50-minute lesson plan on the impact of climate change on coral reefs. Include hands-on activities and discussions."
Keep Refining
Most AI systems remember the conversation's history, so you can give follow-up prompts to refine the AI's output until it meets your expectations.
Ask AI to "Generate a creative writing prompt about a time traveler." You can further refine with "Focus on the challenges of adapting to a new time period."
Prompt Engineering: A Framework
Effective prompt writing, often referred to as "prompt engineering," involves a blend of creative and technical skills. Knowing how to structure prompts is important, but sparking the AI's imagination can lead to more insightful and engaging results.
One popular framework for guiding your prompts is the REACT model, which can be a valuable tool for tailoring your instructions to your specific needs and audience. However, it is important to note that there are many other frameworks and techniques available, each with its own strengths and applications.
R
Role
Clearly define the role you want AI to take as the content creator.
Do you want it to act like an instructor, writer, tutor, coder, designer?
E
Expectation
Clearly define the intended objectives and actions you want AI to perform.
Do you want it to summarize, analyze, create, or translate content?
A
Audience
Define the target audience of the output.
What are their needs, professions, knowledge levels, or demographic information?
C
Context
Provide relevant information to clarify your request.
Include data, templates, course levels, subject matter, and any constraints.
T
Tone
Specify the tone, considering the intended audience and purpose.
Should it be professional, conversational, written in first person?
Examples of Well-crafted Prompts
Let's see the REACT framework in action with some sample prompts.
Sample 1: Discussion Prompt
Role
Act as an experienced sustainability practitioner, specializing in adult education.
Expectation
Design a discussion prompt challenging students to assess their household carbon footprints and identify creative solutions for reducing their environmental impact.
Audience
The audience is adult learners at various stages in their sustainability journeys, some seeking new careers and others looking to advance their current roles.
Context
The online discussion will appear in Week 2 of an 11-week introductory sustainability course. The prompt should encourage Bloom's higher-order thinking, asking learners to analyze, evaluate, and apply course concepts to their own lives.
Tone
Use an encouraging tone, provoking thought and discussion. Keep the writing simple, addressing learners directly.
Click to reveal AI-generated output.
Carbon Footprint Deep Dive
"Welcome to Week 2! This week, we're getting personal with our environmental impact. It's time to take a closer look at your household's carbon footprint.
Your Challenge:
- Calculate: Use a reputable online carbon footprint calculator to estimate your household's annual carbon emissions. Be sure to choose a calculator that is relevant to your location and lifestyle.
- Analyze: Review the results. Which areas of your household's consumption contribute most to your footprint? Were there any surprises? Share your findings with the group.
- Evaluate: Reflect on your emotions and thoughts as you reviewed your carbon footprint. Are you satisfied with your household's impact? Are there areas where you feel you could make significant improvements?
- Create: Brainstorm at least three specific, actionable steps you and your household could take to reduce your carbon footprint. Consider changes to energy use, transportation, diet, and waste management. How could you implement these changes in a way that is feasible and sustainable for your household?
Remember: There are no right or wrong answers here. This is a space for open, honest, and respectful dialogue. Let's learn from each other and inspire one another to create a more sustainable future."
Google. (2024). Gemini (June 12 version) [Large language model]. https://gemini.google.com/advanced
Sample 2: Case Study List
Role
Act as a time-traveling project manager who has witnessed the successes and failures of countless projects throughout history.
Expectation
Generate a list of real-world case studies showcasing the critical role of risk analysis in project management. Projects should span different eras and industries. Include the industry sector, brief description, risks encountered, consequences of those risks, and potential / actual risk management strategies.
Audience
The audience is adult learners in a project risk management course, part of a series of courses to help students pass the PMP exam.
Context
Students will work in groups and select a single case study to focus on for their final research project presentation.
Tone
Keep the case study descriptions simple and present information in comparative table format.
Click to reveal AI-generated output.
Case Study List
"Here are some real-world case studies across different eras and industries, highlighting the impact of risk analysis (or lack thereof) in project management:
| Project Title/Industry | Brief Description | Risks | Consequences | Risk Management Strategies |
|---|---|---|---|---|
| Construction of Great Pyramid (Construction) | Building the largest of the Giza pyramids. | Labor shortages, resource constraints, technical challenges, environment. | Delays, cost overruns, safety hazards, structural issues. | Labor management, logistics planning, engineering innovation, mitigation. |
| Manhattan Project (Military/Research) | Developing the first nuclear weapons. | Technical uncertainty, safety hazards, security risks, ethical dilemmas. | Delays, accidents, security breaches, ethical fallout. | Parallel research, safety protocols, compartmentalization, ethical debate. |
| Self-Driving Cars (Automotive/Tech) | Developing autonomous vehicles. | Technological challenges, safety concerns, regulatory hurdles, public acceptance. | Development delays, accidents/liability, regulatory roadblocks, market rejection. | Iterative development, safety redundancies, proactive engagement, education. |
| Panama Canal (Engineering) | Constructing a canal connecting the Atlantic and Pacific. | Disease outbreaks, geological challenges, financial constraints, political instability. | Loss of life, delays/ cost overruns, project halts, reputational damage. | Disease control, geological surveys, financial planning, political negotiation. |
| IBM Watson for Oncology (Healthcare/AI) | Developing an AI system to assist in cancer diagnosis/ treatment. | Data quality/bias, algorithm transparency, regulatory compliance, ethical concerns. |
Misdiagnosis/ mistreatment, lack of trust, legal/ regulatory issues, reputational damage. |
Data curation/validation, explainable AI, regulatory engagement, ethical boards. |
Prompt Enhancements
While the REACT framework provides a solid foundation for crafting effective prompts, there are additional elements you can incorporate to continually refine your instructions and improve the results:
- Style: Specify the desired writing style (e.g., first, second, or third person perspective), format (e.g., paragraph, bullet points, numbered list, table format), length (e.g., 150 words), and any other stylistic preferences.
- Examples: Provide examples of the desired output. This could include sample text, images, code snippets, or templates that you would like the AI tool to follow or structure the content into.
- Reference Materials: If you have specific resources you want the AI to draw upon, provide links or references to relevant documents, websites, or other materials.
- Source Verification: You can request that the AI tool include sources for any facts or claims it presents to help you verify the accuracy of the information provided.
By incorporating these additional elements into your prompts, you can fine-tune the AI's output and ensure it aligns more closely with your specific expectations.
Additional Prompting Techniques
Beyond the REACT framework and prompt enhancements, there are several other techniques you can use to optimize your interactions with generative AI.
Zero-Shot Prompting
Generating a response without feeding the AI tool any examples or prior context. This technique is ideal when you need quick answers to basic questions or general topics.
One-Shot Prompting
Generating a response based on one example or piece of context provided by the user. This can help AI grasp the intended format and style.
Few-Shot Prompting
Providing multiple examples to guide the AI's response. This is useful for tasks that require a deeper understanding of the context or desired outcome.
Chain-of-Thought Prompting
Providing a sequence of related examples or questions to refine the original question and encourage the AI to think step-by-step through complex tasks or problems.
Self-Consistency
Asking AI to generate multiple outputs, and then selecting the one that is most consistent with its previous responses. This can help improve the quality and coherence of the AI's output.
Template Prompting
Providing sample answers or templates in the prompt itself to guide the AI tool to generate responses in a specific format or style.
Additional Resources
We encourage experimenting with different techniques to discover which ones work best for your specific use cases. The resources below provide additional techniques and sample prompts for inspiration:
Guides
- Prompt engineering (OpenAI)
- Prompt Engineering Guide (Learn Prompting)
- LLM Prompting (Miten MIT)
Videos
- I Discovered The Perfect ChatGPT Prompt Formula (Skill Leap AI)
- The PROMPT Engineering Formula That ALWAYS Works (Educraft)
Samples
- GenAI Chatbot Prompt Library for Educators (AI for Education)
- Prompt Examples (OpenAI)