AI-Based Assignments and Activities – Writing Across the Curriculum

Last Updated: Fall 2025

The most effective AI-inclusive assignments will not only only require students to use AI in some way but actively teach them to consider the ways that AI can and can’t help them to accomplish their goals for a given task. These assignments will help students critically examine the kinds of work required to complete their assignments well, how that work is intended to benefit them, and how AI can help or hinder that work.

Here are a few examples of ways you might do that in your courses:

Modeling Critical Reading/Thinking Skills

Ask the AI to explain a central course concept or summarize a major reading. Then, have students evaluate the results, either individually, in small groups, or as a class.
- Was the AI’s response generally accurate? Did it make up any details or distort any ideas? Does it include questionable data, fictional or uncited sources, or other misinformation?
- What major ideas or important details did the AI leave out of its summary? What seems to be the overall approach it took to summarizing–did it stitch together key words, paraphrase the introduction and ignore the rest, etc?
- Did the answer suggest signs of bias, shallow knowledge, or misunderstanding?
Give students a sample essay written by a real student in response to your a writing prompt for your course. Then, have students feed the prompt into a generative AI tool and have them discuss (collectively in class or individually in writing) how the two essays differ. You could use this as an entry point into a broader discussion of the ways that human and AI writing differ, or simply to illustrate the ways your expectations for student writing might differ from the expectations that drive the AI’s response.
- What are the comparative strengths and weaknesses of each piece?
- What assumptions did the human and AI writers make about the instructor’s expectations, based on the prompt?
- Which essay, overall, seemed to address the prompt more effectively and why?

Developing Metacognitive Skills

Use the AI to analyze an assignment prompt. Have students choose one sentence from the prompt at a time to input into the AI, then see how the results improve as you add more details. Use these results to discuss what the key details of the prompt are and how students can best address them in their writing.
In classes with multiple short essays, have students use an AI tool at some point in their writing process. You could either stipulate a specific use of AI (e.g. “give the AI a prompt and ask it to generate an outline for you”), or simply tell students to use AI in some way that seems beneficial to them. When the students turn in their essays, have them write a “cover sheet” in which they consider how their experience on this assignment differed from previous ones.
- Did the AI actually make the writing process easier? Are they equally happy with the results?
- Do they feel they learned more, less, or the same amount from the experience?
- Do they feel that the results really reflect their own skills and knowledge?
- Do they feel their use of AI would be ethical for an assignment that didn’t explicitly call for AI use?
As a follow-up to that assignment, you might ask students on the next assignment to reverse their use of AI. So, if the students had the AI tool generate an outline, which they turned into an essay, on the next assignment they might write the outline and ask the AI to write an essay based on that outline. Then, once again ask students to write a cover letter for that assignment in which they describe their use of AI and examine how it helped or hindered their writing.

Strengthening Revision and Feedback Skills

Give the students an AI-generated response to your writing prompt and use it to practice peer review. This can be particularly helpful if you also examine a student-written draft in the same exercise. Since AI writing is often grammatically polished but short on content, whereas a human’s first drafts are often rough around the edges but with the seeds of more significant ideas, the students will need to decide how to respond positively and constructively to very different kinds of writing issues.
As a somewhat more intensive version of the exercise above, give students an AI-generated response to your writing prompt and ask them to revise it themselves. Ideally with the explicit caveat that they can’t simply abandon the AI draft and start over. Have students discuss or write about their revision process at the end of the assignment.
- How did they approach the revision?
- What aspects of the original did they keep and what did they change, etc?
- Were there fundamental flaws in the AI’s understanding of the assignment that they simply couldn’t fix or improve?
Finally, as an even further extension of this approach, you might ask students, individually or in groups, to coach the AI to revise its own essay. Give the students a digital copy of the AI-generated essay and ask them to direct the tool to revise the essay according to their specifications. Does the thesis need to be more specific or expansive? Does the essay need to engage its sources more thoroughly? This exercise obviously requires students to generate constructive feedback which is a useful skill in itself, but it also teaches them the limits of the AI’s capabilities. At what point does is the AI simply unable to improve on its own work in a way that’s beneficial to completing the assignment at hand?

Please note that the list of suggestions above is very much a work in progress. To that end: Carleton faculty, if you’ve designed an activity or assignment that incorporates AI writing tools and you’d like to share it with your colleagues, please contact George Cusack.

Ethical Concerns for Creating AI Assignments:

As we explore elsewhere on this site, there are several valid and significant ethical considerations around using AI tools. While we generally believe that the benefits of engaging AI in your classes can outweigh the risks and drawbacks, we also encourage instructors to think through these potential issues so they can make informed decisions on how to proceed.

In particular, instructors should keep in mind that not all AI tolls are created equally with regards to their capabilities and with regards to the ethical issues involved in using them. In particular, the tools that are available to faculty, staff, and students through their Carleton accounts (Gemini, NotebookLM, and Amplify) will not use their interactions with students to feed their training models. With most AI tools available outside of Carleton, anything you or your students upload to an AI tool becomes, to one degree or another, the property of the company that owns it.

It’s also worth noting that teaching students to use AI effectively in your course also teaches them to use this technology more effectively in their future courses, whether their future instructors wish them to or not. Thus, we urge all instructors to address using AI responsibly, not just effectively. Carleton’s Understanding Plagiarism website has a page on AI, which instructors might find useful for discussing responsible AI use with students.

At the very least, though, please reiterate to your students that the boundaries around appropriate AI use will vary significantly from one course to the next, and that each instructor sets the rules that they believe will help students learn most effectively. Thus, it is always to students’ benefit to know what’s acceptable for any given course and assignment and to respect those boundaries.

Practical Concerns for Creating AI-Based Assignments

On a practical level, it’s also worth remembering that the availability of AI tools is by no means equal or guaranteed for all students. All major AI companies require users, at the very least, to create accounts using some personal information, and most also require paid subscriptions to access their most sophisticated features.

The availability of specific tools and features can also change quickly, so instructors could easily plan an activity using a specific tool, only to find that some or all of their students can’t access it when the time comes, or that connecting students to the tool is more difficult and time-consuming than they’d planned.

At the time of this writing (late 2025) the most widely accessible generative AI tools at Carleton are Gemini (Google’s all-purpose chatbot), Notebook LM (another Google tool) and Amplify (a multi-platform tool). All three of these tools are available for free to all faculty, staff, and students. More importantly, Carleton’s user agreements ensures that none of the interactions Carleton users have with these accounts can be used by the companies to train their AI models. In other words, these tools provide both equitable access and protection for users’ privacy and intellectual property.

There are, of course, many (many) other AI tools and platforms available, some of which offer the same basic functionality (with slightly different features and options) as the ones mentioned above, and others that serve a dizzying range of specialized purposes. If you’re interested in seriously integrating AI into your classroom, you should feel free to explore what’s out there, but we encourage all faculty to consult with Academic Technology before they adopt any new tool in their teaching, so they can help you assess what the best tools are for your purposes in terms of functionality, cost, and security.