Teaching with ChatGPT: Assignment Edition

Our previous ChatGPT Teaching Tip offered a broad overview of the pedagogical potential (and pitfalls) of chatbot technology. In this tip, we will take a closer look at some of the most interesting examples of teaching with (or about) the bot shared by educators across America:

Prompt engineering

One common pedagogical use for GPT technology is "prompt engineering": asking students to create and refine prompts for eliciting strong AI outputs. These are skills many students will need in their professional lives, as AI-enhanced work systems become more pervasive. In academic contexts, prompt engineering affords an opportunity for in-depth analysis, pushing learners to consider what exactly makes for a successful code sequence, lab report, legal briefing, and so on. As students coach the bot to improve its outputs—as well as rectify its biases and correct its errors—they exercise precisely the analytical, research, and communication skills we hope to foster in higher education. Fall 2024 UPDATE: while prompt engineering remains a useful pedagogical exercise, many users point out that AI tools can help create and refine prompts.

Example: Ethan and Lilach Mollick share a series of sample assignments involving ChatGPT output analysis and prompt design. In one of them, students ask the bot to add, delete, or refine steps in an explanatory sequence in order to challenge the illusion of explanatory depth. Also from Mollick, a checklist for crafting effective ChatGPT prompts.

Fun with forgery

One sub-category of prompt engineering is getting the bot to produce simulacra of documents (historical, legal, professional, scientific) that could pass for the real thing. We already know ChatGPT can fake plausible scientific abstracts. Students can try their hand at "forging" texts relevant to their disciplines—a legal brief, an 18th century pamphlet, a section of an instruction manual—and then testing their creations on novices as well as experts to refine their understanding of disciplinary norms and genres.

Meta-reflection

One of the most neglected aspects of learning in higher education is metacognition. Our desire to convey the what of our disciplines can sometimes get in the way of pausing to consider the how and the why. The AI revolution affords an opportunity to rethink, together with our students, what skills and knowledges we value and for what reasons. Does assembling and curating pre-written segments of text still count as "writing"? To what degree is relying on the work of others an unspoken part of the creative process and where is attribution still essential? How might the crowdsourcing power of ChatGPT upend our notions of subjectivity and originality? Can AI serve as an equalizer, aiding learners in areas of weakness so their strengths can shine? These questions are inherently interesting to our students, whose futures will unfold in an AI-saturated world. Inviting students to engage in conversations and shared norm-making gives us an opportunity to learn from each other and to start navigating this brave new world together.

Example: Paul Fyfe designed a reflective, human-bot hybrid assignment ("Professor Fyfe's Turing Test") in which students explore questions of authorship, agency, and academic ethics by co-writing a paper with ChatGPT. (See here for professor Fyfe's explanation of his assignment in video form).

Considering the context

It's important to note that many of the above activities require students to create their own accounts with OpenAI. Is this a good idea? ChatGPT presents numerous ethical challenges—from privacy and equity issues to labor and environmental concerns—that may make many professors (and students) hesitant to jump on the chatbot bandwagon. Instead of working with the bot directly, we can invite students to consider the ethical, socioeconomic, and disciplinary implications of ChatGPT ‘s emergence by looking at the bot's key texts (like privacy terms) and contexts. Autumn Caines offers an excellent list of activities designed to help students think through the role of ChatGPT (and AI more broadly) in the contemporary world, from socially annotating the bot's terms of service (TOS) to conducting a techno-ethical audit. These activities allow students to gain a more informed view of AI's implications for their academic disciplines and professional futures without sharing personal data.