With the rise of ChatGPT and other artificial intelligence (AI) programs, a new essential literacy skill is emerging. This skill is associated with the creation and engineering of prompts that users input into AI tools to generate content. We call this prompt literacy. Learning how to write effective prompts will empower learners to be the drivers of AI rather than being driven by it.
When AI is brought into the classroom, whether it is for generating text, images, videos, or anything else, it is critical to note that these tools are prompt-dependent. The better and more nuanced the prompts entered, the more useful and responsive the tools become. As teacher Cherie Shields recently noted on The New York Times' Hard Fork podcast, students are "going to have to know what they're talking about. They're going to have to ask it questions. They're going to have to be very specific."
In other words, students need to learn how to interact with AI programs in the same way they learn writing processes, math strategies, or research techniques. Thoughtful and strategic work on empowering each student through prompt literacy can also maximize the possibilities for personalization. It will equip students to use AI to tailor their learning to their own "needs, goals, and abilities" and to reinforce the concepts and skills they are being taught.
Students need to learn how to interact with AI programs in the same way they learn writing processes, math strategies, or research techniques.
To help educators teach and experiment with prompt literacy, we've developed the CAST model, which is based on effective search engine practices and what we've learned as we navigate this new space. The acronym stands for Criteria, Audience, Specifications, and Testing. A quick overview:
- Criteria refer to the constraints or delimiters for the output. This includes rules or norms that the AI is to follow, such as using short sentences or a bulleted list. Criteria may encompass points of reference, such as place or types of vocabulary. They also define parameters for the output, including the scale, format, and direction, or the type of answer requested.
- Audience refers to who specifically the output should be geared toward. In the sandbox stage, when teachers and students are playing with the tools, the audience is likely just themselves. As they get more sophisticated with prompt engineering, they may want to ask the AI to produce something for a particular audience, such as 4th grade students, a group of expert meteorologists, or a committee on diversity and equity.
- Specifications for prompts are where the user includes relevant details and descriptions. For example, users may input specific examples for the AI tool to emulate, or they could give the AI context to situate a particular result. In our view, specification-making is one of the highest forms of prompt literacy. Its utilization points to richness of meaning and higher-quality results. Good specifications are the difference between "Write a poem about summer" and "Write a poem in iambic pentameter using a languid tone." Another example would be a student pasting their own writing into ChatGPT and asking the program to analyze the text for grammar, voice, tone, or sophisticated language. Consider what a big upgrade this is from tools like Grammarly or the built-in features of Microsoft Word or Google Docs to catch general mistakes.
- Testing is perhaps the most important factor in the development of prompt literacy. Here, students iterate and refine their language to build their knowledge of what AI can do and how they can make it work for them. They can add new keywords to their prompts, revise their original prompt with different parameters, ask the AI to help them brainstorm, or lay out a logical sequence of steps. With testing, students gain agency and reflective skills within AI environments. To support students in this work, ask them to explain how they built their prompts, tweaked them, and ultimately how the final information or product was deemed satisfactory for their work.
As part of prompt-literacy development, educators should also encourage students to defend or extend their output. Why is this the best answer? What products could I create with this information? Did the AI get it right and how do I know? This kind of analysis of the content AI generates in response to their prompts is essential to students' understanding and constructive use of AI-based technology.
An Evolving Pedagogical Space
AI prompt literacy is an evolving space where educators can explore new pedagogical possibilities and better prepare their students for the future. Ultimately, supported by models like CAST, it can help personalize learning experiences, provide opportunities for more immersive and interactive learning, and enhance students' creativity and ability to build products that have never before been possible. It can also help students develop skills like corroborating and thinking critically about information—skills that will become increasingly important in the age of AI.
