Artificial intelligence (AI) is shifting how we experience reading and writing. Students and instructors alike need some understanding of the main limits and affordances of AI-generated writing. This page offers some basic information for writing instructors and some guidance on how to talk with students about generative AI.
Large language models (LLMs), like ChatGPT, use machine learning algorithms that produce responses to user-generated prompts and mimic human communication patterns based on the input they receive. ChatGPT and similar tools (like Gemini, Copilot, and Claude, etc.) use a chat-based environment that allows users to engage the tools in conversation on topics of their choosing. Users can ask questions, provide prompts, input large chunks of text, attach PDFs, and request specific actions (summary, proofreading, etc.).
Drawing on massive datasets that include billions of bits of text, LLMs can generate output reflective of a range of genres, including the academic essay. The size of those datasets and the architecture that houses them will very likely improve over time, including the tools’ ability to generate language that approximates human expression.
While there are plenty of reasons to be suspicious of AI-generated writing, we recommend a classroom approach that strikes a critical, reflective balance and supports learning objectives regarding such issues as engagement with sources, information literacy, genre awareness, and others. By addressing generative AI directly and being explicit about how and when it can be useful in the composing process, instructors can help close the “AI literacy divide” in a way that supports our equity goals as well as our learning objectives.
Teaching Recommendations
Before the term: We strongly recommend that you familiarize yourself with at least one generative AI tool to understand the kind of output it produces relative to your course topic. It may be helpful to explore how ChatGPT or Copilot (the Pro subscription version of which is now free to all members of the BU community) might help you in generating or revising essay prompts about your topic, creating classroom discussion questions based on course readings that you share with it, suggesting examples or applications of course concepts that are relevant to your students, or suggesting and summarizing new course readings. Once you find some ways GAI tools–in combination with your expertise and critical judgment–can be helpful to you in your teaching, you may arrive at new ideas for how it can be useful to students in their learning as well.
It is not necessary to cite the AI tool for these two Writing Program-authorized uses. If I authorize additional uses of AI tools for particular assignments, I will do so in writing, and I will offer guidance on if and how to cite the AI tool (MLA, APA, Chicago). Remember that ultimately you are responsible for any work you submit, including any language, ideas, or information that is introduced through your use of AI tools.
This resource may help you think through the (complicated!) questions of academic integrity across a range of possible student GAI uses. This flowchart may help you define an approach to GAI that supports student learning in your own teaching context.
At the beginning of the term: Within the first two weeks of the semester, set aside time in class to talk with your students about AI-generated writing. Students are likely using GAI in a wide variety of ways for academic and other purposes, including ways that you are very unlikely to detect. They want and need guidance from their instructors about which uses are acceptable and clear explanation of why.
- Ask students about their knowledge of and experiences using AI tools. Consider incorporating Five Things Every College Student Should Know About AI-assisted Writing. Communicate the Writing Program’s baseline AI use guidelines above (and included in Section 2 of the WR syllabus templates). Make sure they know that different classes at BU may have different expectations about what kinds of AI use are permissible.
- Remind students that text they enter when using LLM platforms may no longer be private. It will be incorporated into the greater pool of data that the LLMs draw on, fair game for reuse without their permission and reason not to input sensitive or personal information. (One advantage of using CoPilot Pro through a BU subscription is that privacy protections are built in.)
- You may also want to incorporate some of the following:
Chat GPT 4-o prompt: “Generate an image of a typical BU classroom”
-
- In small groups, ask students to choose a freely available GAI tool and use it to generate images of something that the students are sure to know a lot about–for example, a class at BU or some BU students. Then come together to share and analyze the different results. The images will likely be some combination of impressive, trite, biased, and inaccurate, depending on the tool and the prompt. Though image generators and text generators don’t work in exactly the same ways, using images as examples can help you to quickly raise important questions about GAI’s powers and limitations.
- Highlight a particular rhetorical situation and consider whether AI output is reliable or useful given the context. How would you regard a thank-you note generated entirely or partially by AI? How about an instruction manual, a recommendation letter, or a short story? What’s a situation where a text’s originality doesn’t matter at all? What’s a situation where AI’s inability to feel or make ethical judgments matters greatly? What’s a situation where AI’s capacity to generate large numbers of options is an asset?
- Identify examples of the limitations of inexpert or carelessly used AI-generated content: Misinformation, misattribution, reproduction of bias, generic style, vague/superficial content
- Consider how generative AI might be helpful in the writing process: Brainstorming, outlining, generating counterarguments, offering formative feedback, standardizing format and language
If you plan to authorize uses of AI tools that go beyond the Writing Program’s AI use guidelines, help the class come up with a class agreement that details expectations and terms of ethical and responsible use for your section’s assignments, perhaps as part of a larger classroom community agreement discussion. Put detailed class AI guidelines in writing and offer guidance on citation.
Throughout the term: Continue to emphasize writing as an experience and a process, supported by in-class scaffolding activities that engage students in critical thinking, collaboration with peers, and reflection. You may want to revisit class expectations for ethical and responsible use of generative AI later in the term, in the context of particular assignments or activities. If you authorize particular uses of AI tools, do so explicitly and offer students guidance on citation. Avoid using AI tools that charge fees, which may put an unanticipated financial burden on students.