August 27, 2024
This is the first in a series of posts about AI strategies in teaching and learning.
The recent Walton Foundation study on AI in education reported that 73% of private school teachers in the United States had received no AI training. Even now, 19 months after ChatGPT arrived, many independent schools are struggling to form AI policies to guide educators and students. This lack of guidance underscores the urgent need for AI education for educators.
One of the most important skills for educators in the age of AI is prompt engineering — the art of crafting effective prompts to elicit desired responses from AI chatbots like ChatGPT. While AI is getting better at understanding our requests, the quality of an AI chatbot output directly depends on the quality of the input, making prompt engineering a foundational competency for teachers looking to maximize the potential of AI for instruction.
Context-Principle-Instruction (CPI)
To help educators harness the power of ChatGPT and other AI tools, I’ve developed the Context-Principle-Instruction (CPI) approach to prompt engineering. The CPI framework emphasizes providing relevant course context, adhering to research-based principles for generating effective prompts, and offering clear, affirmative instructions to guide the AI’s output.
The “C” in CPI stands for Context. When crafting prompts for instructional purposes, include the course name, grade level, proficiency level, curriculum unit, topic description, and length designations. This information helps the AI generate more relevant and targeted outputs. The name of your course is especially important if you teach a standardized course, such as Spanish 1 or an Advanced Placement course. AI has access to many resources for standardized classes, so if you fail to mention the name of your course, you may miss out on helpful information.
The “P” represents Principles. Apply research-based guidelines for creating effective prompts, such as making the task clear, considering the intended audience, requesting detailed responses, asking the AI questions, using example-driven prompting, and working to reduce bias. For example, if you don’t ask for a detailed response or specify the number of words you expect, the AI may provide a briefer response than you anticipated.
The “I” signifies Instructions. Focus on affirmative instructions while carefully considering and iterating on the prompt’s wording. Clear, specific instructions lead to more accurate and useful AI-generated content. For instance, focus on telling the AI what you want it to do, more than telling it what you don’t want it to do.
Here is an example of a CPI-guided prompt:
Prompt: “Create a 6th-grade social studies lesson plan for a 60-minute class on South America’s geographical features and climate. Include learning objectives, key vocabulary, an engaging activity, and an assessment. Ensure the content is appropriate for sixth-grade students and aligned with Common Core Curriculum Standards.”
Explanation: This prompt is effective because it provides clear context (grade level, subject, and topic), specifies the desired output (a lesson plan), and includes key components (objectives, vocabulary, activity, and assessment). It also ensures the content is age-appropriate and aligned with educational standards.
Prompting Strategies and Libraries
Several effective prompting strategies can significantly improve the quality and accuracy of AI-generated outputs. I often use The Flipped Interaction Strategy, which involves asking the AI to pose questions before providing a response, enabling it to gather more precise and tailored information. I also prize The Reflection Strategy, which entails asking the AI to explain its reasoning and assumptions, providing valuable insights into its suggestions, especially when exploring less familiar topics. Also helpful is the The Question Refiner Strategy, which allows the AI to refine your initial prompt by suggesting an improved version of your prompt.
Furthermore, helpful AI for education prompt libraries, such as The AI for Education Prompt Library, The AI Prompt Library, and Ethan Mollick’s Prompt Library, offer templates and examples organized by category, subject area, or educational purpose. These curated collections provide a solid foundation for teachers to build upon and customize for their specific needs.
AI Chatbot Memory and Personalization
Even so, if there’s one truism in AI prompting, it’s that you will likely have to refine your prompt. It’s to be expected, but it’s also a hugely beneficial process. For one, ChatGPT becomes more helpful as you converse with it, continually remembering details and preferences. AI’s probing capabilities include the ability to ask follow-up questions or request clarifications. ChatGPT can also carry what it learns between chats, allowing it to provide more relevant responses.
You can help yourself by personalizing ChatGPT. It offers Custom Instructions, whereby you can customize your interactions with ChatGPT by providing specific details and guidelines for your chats. You can also manage ChatGPT’s Memory by indicating what you want it to remember and what you don’t want it to remember.
Consequences of Shying Away from AI Prompting
Many educators are shying away from AI prompting. Over the last year, teachers have been gravitating towards AI teaching assistants such as MagicSchool, SchoolAI, Brisk Teaching, and Eduaide, which offer streamlined educational AI tools. These assistants minimize the need for effective prompting, but they often pale in power and flexibility to ChatGPT. Many AI teacher assistants run off a less-powerful large language model than ChatGPT’s GPT 4o (Omni), arguably the world’s most advanced and powerful LLM.
There are immediate consequences. For instance, GPT 3.5 is much worse at math than GPT 4o. ChatGPT’s flexibility and adaptability also allows it to engage in a broader range of topics and adapt to diverse educational contexts and needs, often surpassing the capabilities of more specialized AI tools. This flexibility enables educators to explore lesson ideas across various subjects and grade levels without being limited to the specific tools or templates that specialized education AI tools might feature.
An Iterative Process
Remember that AI prompting is an iterative process. Experimenting with different prompts and observing the AI’s responses will naturally lead you to refine your techniques and develop a style that works best for you and your students.
Embrace the challenge of mastering the art of AI prompting, and unlock the power of AI to create more engaging, effective, and inspiring learning experiences for your students. Dive in, experiment with prompts, and discover how AI can enhance your instructional design, taking it to new heights.
Tom Daccord taught in independent schools for 15 years and is currently an AI speaker, trainer, and author. His most recent book “AI Tools & Uses: A Practical Guide for Teachers” is available at Amazon.com.
You may also be interested in reading more articles written by Tom Daccord for Intrepid Ed News.
