The KNIGHTLYnews is an online forum where FWS instructors and other teachers of writing can swap and share ideas for best classroom practice. Weekly posts are designed to help teachers develop lesson plans and writing assignments, and respond to classroom challenges by introducing new teaching tools and sharing emerging pedagogical ideas. Posts also direct readers to program and campus resources that support teaching and learning, and provide opportunities for peer collaboration and mentorship.
As generative AI has become ubiquitous, writing instructors are being asked to guide students on how to use these technologies effectively and ethically. The Knight Institute’s AI policy and guide on AI syllabus statements offer helpful starting points. Yet, with numerous suggestions and a growing list of AI tools, it can feel overwhelming knowing how and when to incorporate AI in writing pedagogy. Many of us fear that AI will distract from our core teaching goals, including supporting students to develop their own voices and ideas.
In this post, I’ll suggest that in order to decide whether and how to use AI in the classroom, we should start with our central teaching goals. What writing skills, techniques, and dispositions do we want students to build in our class? Being firm in these goals has helped me decide how to use AI in ways that support this learning. In the rest of this piece, I outline three broad reasons you might want to include AI in your writing course. I then present a class activity that employs AI to help students practice a core writing move.
AI for Teaching Writing: Three purposes
Students can and do use AI for a range of writing tasks, including brainstorming ideas, and drafting, analyzing, organizing, and editing texts. As you consider how to incorporate AI in your activities, I suggest you begin with this question: Do you see AI as itself a valuable writing skill, or do you see these technologies as simply sources of texts ad other kinds of learning materials? In other words, do you want students to learn to use AI as an important new writing skill, or do you merely want to use this technology and its outputs to help students to practice familiar, “analogue” writing skills? How you answer this question might lead you towards one or three three possible purposes for teaching writing with AI:
AI as a writing skill/tool: Do you believe that AI can contribute valuably to the writing process? If so, you might want to intentionally teach students use these tools effectively and ethically. Here, using AI is itself a learning goal. For example, you might want students to learn to use AI to brainstorm for essay topics, or to edit their drafts to remove grammatical errors or “written accents.”
AI as a source of learning materials: Alternatively, you may be committed to teaching analogue writing skills, and view AI as largely superfluous. Totally okay! But AI might still be useful to you for generating sample prose and ideas for students to analyze, critique, or correct—in service of learning analogue writing skills. In this approach, students could even use AI-generated text in the same way they would work with sample text from previous student essays or scholarly works. For example, they could revise AI-written sentences for clarity and brevity, or use AI to generate counterarguments that they might respond to in their essays. Because AI can create and modify these samples on demand, they might be especially useful to new instructors who can’t have a corpus of previous students’ writing to draw on.
AI as a topic of inquiry: Neither instructors nor students know enough about how AI performs writing and what tasks it does well. The large language models from which specific AI systems are built are black boxes. And our limited understanding may prevent us from giving specific guidance to students. Having students experiment with them can be one way to learn about their capabilities and limitations, and to establish shared classroom expectations.
These are the three general purposes for including AI in your classroom: AI as a skill, as a source of materials, and as an object of inquiry. Knowing which of these is your primary purpose—and how this relates to your course goals—can help you to design useful and targeted activities.
For example, if your goal is to use AI to generate learning materials, you don’t actually need to create wholly new, AI-centric activities. You could simply incorporate AI component into an existing activity. The example below models one way to do this.
Example Activity: Using AI to practice summarizing
This activity was part of a learning module in which students learn to summarize scholarly texts. The goals of this unit are for students to: (1) paraphrase dense prose using familiar and manageable language; (2) summarize at different scales, from explaining an author’s purpose to depicting a specific moment in the text; and (3) use paraphrasing to advance an argument about the implications of a claim in the text.
The AI-based activity came in the second of this module’s three weeks. Students had already read the source text and written “analogue” summaries of several sections. In this activity, I asked them to use AI to create summaries of the entire chapter. They then discussed, evaluated, and revised these summaries.
Activity: Revising an AI summary
Students worked in groups of two or three to create summaries of the chapter through the steps below. After composing their summaries, several groups presented their assessment of the AI’s strengths and weaknesses in summarizing, and they explained why they had made the revisions they did in their final summaries.
- Use ChatGPT to create the best possible two-sentence summary of the chapter. Only use prompts. Do not manually revise the output. Save your dialogue with the AI as part of your submission.
- Identify strengths and weaknesses of the AI’s summary, based on our discussions of qualities of effective summary.
- Write your own, improved summary. You are welcome to borrow language or organizational features from the AI-produced summary, or you may write a summary that is entirely your own.
The box below compares one of the GPT-generated summaries with this group’s revision. Note that the students maintained a significant portion of the first two paragraphs, while swapping out some specific terms (“monocultures” instead of “same trees”) and switching to a more active voicing of the first sentence (the subject is “German foresters” instead of “scientific forestry”). The students also removed the AI’s final sentence, which was a hallucination that brought in material from a different chapter of the book (which students had not read). In its place, they wrote their own summary of the overall argument of the chapter.
|
GPT-generated summary |
Students’ summary |
|
Scientific forestry, developed in the 19th century, simplified forests by planting the same trees in rows and organizing them uniformly. However, this led to ecosystem collapse and biodiversity loss. Overall, this approach reflects a broader problem of decision-making ignoring local knowledge, which is often more attuned to the complexities of the local environment. |
German foresters developed scientific forestry in the 19th century. They createdsimplified forests by planting them as monocultures and organizing them uniformly. Although this produced more timber in the short run, it eventually led to ecosystem collapse. Overall, Scott argues that these failures were due to foresters’ narrow, utilitarian mindset. |
Lessons
By analyzing and revising the AI-generated summaries, students built important skills for writing summaries. They had to assess the emphasis of a passage, compare and evaluated alternative wording, and revise sentences for clarity and directness.
Through this process, we also learned about ChatGPT’s capabilities as a summarizer. Students found that, while the AI could readily condense long passages, it often produced a mere list of topics without a sense of emphasis. Others pointed out that ChatGPT—like many human writers—struggles to change its approach. Despite prompting, it would not change the key terms or sentence structures from its first draft. Finally, GPT often “hallucinated,” adding ideas to the summary that were not present in the chapter we read.
By discussing their findings, the class gained a better understanding of features of AI summaries, and proposed some potential strategies for how to use these tools as part of their own writing process (for example, generating alternative wordings of key concepts that students could borrow from). In this sense, while my original purpose had been to use AI to create materials, we ended up learning about how it works (AI as inquiry) and identifying strategies for using it appropriately to support summarizing (AI as a skill).
I hope these ideas and activity are useful for you. If nothing else, the activity to revise AI summaries models how simple it can be to incorporate AI into an existing activity. Experimenting with AI in such controlled and focused ways has helped me to clarify my broader learning goals, course AI policies, and the guidance I give students.
Ewan Robinson
Visiting Lecturer, Knight Institute for Writing in the Disciplines
PhD Candidate, Global Development
