Prompt Engineering for GenAI Tools
Written by Omelia Tennant and Jill Ballard
March 11, 2024 • 2 minute read
What is prompt engineering?
To harness the potential of GenAI tools, the ability to craft effective instructions to elicit relevant, desired generative output is a crucial skill. This can range from a single query to more complex sequences of requests.
Why is prompt engineering important?
Anyone using a GenAI tool is a prompt engineer. Yet developing the ability to craft more effective prompts to elicit relevant generative output is a crucial skill—as well as a foundation of AI literacy. This is often an iterative process of refining prompts to reach targeted output. For instructors, getting more familiar with GenAI tools and improving prompt engineering skills will help to inform decisions about GenAI tools’ use in the classroom as well as to support students’ developing prompt engineering skills.
Prompt engineering is an iterative process beginning with an initial prompt followed by additional prompts to refine a desired output. As with any skill set, there are a range of simple to advanced techniques, which can be employed.
For GenAI text generation tools such as ChatGPT or Perplexity, basic prompt engineering techniques include:
- Be specific — start with a prompt that defines a clear focus, such as “create a discussion post on [add topic]”, to produce a better initial output.
- Add context — include additional criteria for more distinct output, such as “create a discussion post on [add topic] requiring comparative analysis for undergraduate students”.
- Provide examples — for better definition, indicate a specific format, provide descriptive text, or add an existing example such as, “create a discussion post on [add topic] for undergraduate students using the following [add discussion post example].”
Creating prompts for GenAI image generators is similar and will also require iterative prompting to generate a desired output. Starting with a simple prompt, such as the word “dog”, will generate an image, but adding specific and contextual details, such as “a goldendoodle wearing pink sunglasses in the arctic”, will produce a more complex image.
For more advanced prompt engineering methods using text generators, try these techniques:
- RTRI method: this process involves creating a role, task, requirements and instructions to prompt a GenAI tool generation.
- TAP & TASTE method: this two-step process involves setting the parameters for prompts and then critically evaluating the generated content. This method can be used to support students to develop prompt engineering skills.
As a reminder when creating prompts, never input personal information or university data classified as internal, confidential, or highly confidential. Refer to the USF’s Guidelines for Responsible Use of Artificial Intelligence (AI) to learn more about safer GenAI tools use.
Everyone who uses a GenAI tool is a prompt engineer. Here are some examples of how educators are prompt engineering:
Practice your prompting skills with GenAI text generator ChatGPT.
To learn how to use ChatGPT and other GenAI technologies, schedule a training session with Instructional Technologies & Training and/or schedule an Instructional Design consultation.