Prompt engineering is an important part of using GenAI. In order to maximize our use of these tools, we want to input our questions or requests (referred to as a prompt) in the best way possible (creating or engineering the prompt) to get the best responses. I gave ChatGPT the following prompt: Provide a few sentences to explain prompt engineering for generative AI to a novice. CHatGPT gave me the following response:
“Prompt engineering is like giving specific instructions to a smart computer program. Instead of just asking a question, you carefully design the way you ask it to get the best results. It's about fine-tuning how you communicate with the computer, making your input more precise and tailored, so you can get the information or output you want. Think of it as crafting the right question or request to help the computer understand and respond in the most helpful way.”
There are several models, formulas, frameworks, and suggestions for how to craft prompts. Many of these models are very similar and provide great guidance for prompt engineering. It is important to note that these models haven't been scientifically tested, meaning we can't point to one and say that it has been proven to be more effective for writing prompts. Instead, we should think of these models as guidelines. Be thoughtful and strategic about how you write prompts. Learn from each interaction about how GenAI tools work and apply that understanding to your next prompt.
With that said, here is a simple model that can help you to craft effective prompts. The CAST model was developed by developed by Heidi Jacobs and Michael Fisher of Association For Supervision of Curriculum and Development. It contains 4 elements: Criteria, Audience, Specifications, and Testing. Usually, it is best to include all of these elements, but not every request or input may require that. Think critically and use your best judgement. Part of working with GenAI is trial and error.
When writing your prompt, first think about the output you want to receive. The parameters that you set for the output is the criteria. Parameters may include the output format, length, or other details.
Your prompt should indicate who the output is for (the audience) and what their level of understanding of the topic is. This will make the output more relevant and tailored to the user. In some cases the audience may be you, but in other cases you may want AI to generate information for other people. Depending on the topic, the intended audience may have a different level of understanding or background knowledge than you, so it is important to specify this within the prompt.
Prompts should include as many relevant details as possible along with specific examples if needed. Generally speaking, the more detail you provide, the more relevant the output will be. Include specific examples of topics that you want the output to address or include if that is important to your request.
Although we may follow the guidelines for effective prompt writing, sometimes we don't get the result that we need or expect. However, we can learn from each response to help us improve our prompts for next time. Each interaction with Gen AI can help us better understand how the tool works and how we can best structure our prompts to get the best result. Testing includes analyzing, revising, and reflecting.
Reflect: What did I learn about how AI works that I can apply to my next prompt?
Looking to level up your job applications and interview preparation? Enroll in the self-paced Canvas course: Next Level Job Applications and learn how to use Gen AI to enhance your resume and promote your best self to employers.