Five Ways to Design Assignments for the AI Era

As generative AI rapidly becomes mainstream and its use by students for their writing assignments grows, teachers need to be prepared to ensure their students can still get the most out of their work.

“Teachers need to decide when it can and whether it should be used at all, and anticipate how students might use it,” says Amy Stornaiuolo, a professor of literacy education at Penn GSE and an expert in digital literacies. “Because if you don’t decide in advance as a teacher, students might not always think through the ramifications of it and might use it in ways that you did not intend or anticipate. And that can make the entire process unsatisfying for everybody.”

That’s why she created a framework for AI integration in writing instruction developed through the Digital Discourse Project.

“The idea is to decide what your intention is for an assignment—what do you want students to get from it, what do you want to be able to learn from students about their learning—and then use that to decide how and whether AI should be incorporated,” she says.

The framework provides five categories for handling AI in writing assignments. Educators can even adjust their approach for incorporating—or dissuading—AI usage from assignment to assignment based on their goals.

Stornaiuolo provides the following advice for each of the framework’s categories.

Assistive: Supporting the practice of writing

Assistive AI usage in assignments can be as simple as employing grammar checkers like Grammarly. But Stornaiuolo says students may benefit more from complex and underutilized uses, like AI in the role of a “teacher on demand” or thought partner.

When students have difficulty understanding a teacher’s feedback—for example, a note that a passage in their work is unclear—they can input the passage and feedback into an AI tool to unpack why it might lack clarity and provide advice on how to improve it.

Stornaiuolo also points to a tool created by colleagues through the National Writing Project called Thinking Partners. It helps students create tutor or partner GPTs to ask questions or bounce ideas off, and it even includes premade GPTs that assume the role of some authors.

“This category provides teachers a really wide range of potential uses to suggest or avoid,” she says. For example, a teacher may tell students they can use AI for brainstorming or for feedback on their writing, but not to actually generate any text, depending on what is most compatible with the teacher’s goals.

Resistive: Foregrounding student ideas and voices

The resistive category is all about designing assignments that make AI use difficult or impossible. Some of the more obvious resistive methods include completing certain assignments in class with pen and paper and/or timing them.

But, Stornaiuolo says, educators can get creative in designing assignments to dissuade AI use. This includes work that is highly localized to issues specific to their own community or personalized to students’ own experiences, documentary-style assignments requiring interviews of local people, or hands-on activities. Sample activities might include having students take home sticky-note annotations from class to incorporate into their analysis or record a website walkthrough and incorporate their observations into their writing.

Creative: Exploring expressive writing dimensions

In this category, students actually use the AI as a creative partner in their writing. Stornaiuolo says many of the high school writers she works with will program a GPT to act as a character in their stories and then ask it questions to help build the character’s backstory. In this way, the AI becomes embedded in their world-building practice. One student even wrote a story that featured AI as a character, and had the AI write its own dialogue for the piece.

One teacher Stornaiuolo worked with had students research authors and write bios for a public website. They then used an AI image generator to create an image to accompany the bio. With the assignment, the students had to turn in an explanation of the prompts they used to generate and tweak the image, and why they used those prompts. This added a resistive element to this creative use—the students couldn’t rely on the AI to explain their own reasoning behind their prompts.

Rhetorical: Highlighting how writing works

Rhetorical usage helps students analyze writing itself, like teaching them how to change tone for different audiences, assess the accuracy of information or quality of sources, and think about their own writing strategies. Examples of these assignments often involve asking students to generate a text with AI and then annotate and critique the output.

A teacher might ask students to prompt the AI for information about something factual, like a historical event, and analyze the output for accuracy and bias using their own sources. Alternatively, students could compare one generated text to another or even to their own writing to review rhetorical differences.

Critical: Learning how GenAI technologies work

Assignments that analyze AI itself, particularly for ethical concerns, fall under the critical category. This also allows students (and teachers) who might object to using AI for ethical reasons to choose whether or not to use it. An assignment might be as simple as having students research and write a paper about a certain aspect of AI, like its environmental impact, bias, or even the accuracy and biases of AI detectors.

Another assignment might be to present three texts and ask students to identify which one was produced by AI, explain why they think that, then write about what that says about the nature of the technology and how it works.

***

Stornaiuolo points out that many assignments incorporate multiple categories. “Layering these together can be really powerful,” she says.

Teachers should still refer to their district’s or school system’s policies when incorporating AI into assignments, Stornaiuolo cautions, but the framework can help them implement it within those guidelines or in the absence of them.

Above all, Stornaiuolo hopes this framework can help promote positive student–teacher interactions. AI provides students the opportunity to offload their thinking to it rather than putting the work and effort in themselves if they don’t see the value in an assignment.

“Teachers want students to do their own thinking,” she says. “So, instead of framing this as a policing mechanism, this kind of framework moves us away from positioning teachers and students in adversarial roles and toward finding common purpose and being intentional with their goals.”