Introducing Prompt Engineering Using Game-Based Learning

For context, when I first started teaching comp/rhet, I realized fairly quickly that students had very little awareness of the "higher order concerns" (HOC) found in effective writing (i.e. Focus, Audience/Purpose, Organization, Development). To bridge this gap of understanding, I created a game to teach each HOC. Not in my wildest dreams did I think I would use the Audience & Purpose game to teach prompt engineering.

The game is fairly simple: using a website that randomizes words you submit in list-style, I provide students with one randomized genre, one randomized purpose, and then we move through usually 3-4 rounds using different randomized audiences. So, for example, they may get "Text message," "Apology for missing an important event," "loved one," for round one, and then we stick with the format being text message style writing and the purpose being to apologize but the audience might change to boss, or someone they hate but have to put up with, etc.

Students get a practice round in which they can pause, ask questions, etc. Then, each round after that is timed and the audiences get progressively more difficult as they continue. Originally, students used paper to complete each round, then, a Google doc, and now, because we use chatbots, I made the randomized lists even more challenging by adding more specific requirements like "add tone," "must be written in X style," etc.. Once the rounds are done, we compare and contrast how each one changed in tone, word choice, length, and so on.

The goal is for students to see, although genre organizes their message, their purpose should be informed by their audience so they can make more intentional writing choices; however, the new version of the game allows students to understand what an effective prompt requires (which is essentially the same thing effective writing requires). Students still receive the same instructions as above, but, with the help of one of my other amazing coworkers, I was able to modify it to teach it in a less content-heavy approach.

The framework we used for the prompt itself was very basic: a Role for the prompt, a Description of the context, the Specific output, and then Format. Students would receive the purpose from the game, as well as the audience being targeted, but the rest was up to them to prompt the bot. Many students inherently realize that one prompt was not going to yield the best result, so they continued trying to revise it. The more they compared and contrast the results, the more interested they became in the differences between Claude, ChatGPT, and Bard's outputs and each other's prompts.

It was an incredible experience to see how prompting the chatbots pulled out some amazing creativity from the students. On top of that, Claude definitely stepped up my game when it came to creating a longer, more diverse list, of genres, audiences, and purposes. AI continues to amaze me as I continue to improve upon the materials I have built for students to be excited about engaging with.

More to come soon!

-Kayla