If you've been following the news, you've probably heard about ChatGPT, a language model that is able to draw upon artificial intelligence to compose snippets of essays, computer code, and other work that may be hard to tell apart from authentic student work. If you and/or your colleagues or teaching staff would like to have a conversation with someone at the Bok Center about how to approach assignments and academic integrity in this new era of artificial intelligence, please reach out!
In the meantime, here is some advice about how you might adapt your teaching to account for the challenges and opportunities posed by the new technology.
What do I need to learn about generative AI?
Instructors should experiment with generative AI tools to see how they respond to course assignments, consider what students can learn from using the tools, and anticipate the challenges they may pose. Specifically, we would suggest the following course of action:
-
Make sure you are acquainted with the university’s latest guidance about responsible generative AI use, including policy and privacy considerations.
-
Read through the Bok Center’s Introduction to Generative AI.
-
Create account(s) on a few of the most common AI platforms supported by the university and familiarize yourself with their user interfaces.
-
Visit the AI Pedagogy Project, created by the metaLAB at Harvard, for a guide to getting started with AI tools, an LLM Tutorial, additional resources, and curated assignments.
-
Try using AI tools to complete some of the most common/most important intellectual tasks that you expect of your students. (You may be able to enlist colleagues and/or other members of your teaching staff to participate, as well, as many instructors are interested in learning together what the capacities and affordances of these tools are.) This might include things like:
-
Explaining or illustrating a foundational course concept.
-
Identifying a good research question or topic.
-
Generating an argument or outline for a document (essay, lab report, etc.) of the kind that you expect students to write.
-
Drafting, or enhancing, fragments of prose of the kind that you expect students to write.
-
Imagining how someone holding a different identity or position from their own might react to a given argument or set of circumstances.
-
Solving a practice problem or drafting code (or obtaining models or hints about how to do this).
-
Producing multimedia presentations, images, etc. of the kind that you expect students to create.
-
How should generative A.I. affect my teaching and the work that students do in my course?
Once you have a sense of how a student might use A.I. to complete some of the tasks that you might ordinarily assign them, we suggest that you reconnect with your goals and values with regard to what you want students to learn in your course. Does A.I. leave some of them unchanged? Render some of them moot? Allow you to scale up or drill down on some?
-
Reviewing the Bok Center’s course design resources can help you get in touch with the goals for the course, and help you decide where you might wish to “defend” your objectives against AI vs. where you might be content to evolve your objectives, or even lean into AI, now that you have seen how students might use it.
-
Not all assignments are teaching and/or measuring the same skills. Do you assign your students an essay in order to see them produce a polished piece of scholarship, or because essay-writing happens to be one of our best technologies for externalizing our inner thought process? Is it more important for students to show you that they can complete a series of actions successfully, or is the point to see what they are thinking while they attempt to complete those actions? Depending on what you are trying to learn about your students' knowledge and abilities, you may be more or less comfortable allowing them to incorporate artificial intelligence into their research / writing / coding / composing / editing process.
-
If your course typically features a number of writing assignments, ranging from weekly reading responses to a final research paper, you may find helpful advice about how to augment or revise your writing prompts in the Bok Center's Guidance on Generative Artificial Intelligence and Writing Assignments, as well as the Harvard College Writing Program's Framework for Designing Assignments in the Age of AI.
-
If you are contemplating moving towards more in-class assessment (i.e. blue book exams or similar, seated assessments), you may find helpful advice about how to stage them in the Bok Center's Guidance for Instructors Moving to Seated Exams.
How do we develop a policy statement for the course syllabus?
With an eye to what you’ve learned through your experiments with generative AI, you should draft a course policy statement that speaks clearly to students about the ways in which AI may and may not be used in the context of your course. You may find it helpful to review the syllabus statement advice provided by the Bok Center and the Office of Undergraduate Education.
How will I know if my course policy is complete?
As you draft and revise your course policy, you may wish to use the Bok Center’s Illustrated Rubric for Syllabus Statements about Generative Artificial Intelligence as a checklist to make sure you are anticipating the full range of things that students may want to know about your policy.
How should we communicate the course policy to students?
While incorporating your policy statement into your syllabus is a good start, it is likely not the only way that you will want to share it with students. You may also wish to …
-
Post the statement prominently on your Canvas site
-
Reiterate the statement on each assignment prompt
-
Discuss the policy in course meetings
-
Designate occasional office hours for students to ask questions about how to interpret the policy with reference to specific assignments
Ideally, your statement of course policy will function less as the “last word” on AI use, and more as an invitation for students to come forward and share their questions and thoughts with you as we figure out, together, how to apply AI in responsible ways that enhance learning. You may wish to set the expectation early in the semester that the policy may be updated as new use cases are discovered.
How should we enact the policy and apply it to the in-class work and assignments in the course?
You should talk regularly with the teaching team and with students about how your policy will be applied throughout the course, making reference to how it might affect specific assignments and coursework. The other members of your teaching team (TFs, TAs, and/or CAs) should understand clearly the procedure/chain of command in the event that they are confronted by a question or possible policy violation. Should all student questions about appropriate AI use be escalated to the course head? Where should they go with their own questions? What should they communicate to students in the event that they are unsure of how to respond in the moment? Who will be responsible for issuing updates to course policies, if/when such updates become necessary?
May I use generative AI to assist me in grading and providing feedback on student work?
It is unlikely. Sharing student work with a third-party platform (e.g. by uploading student work into ChatGPT so that it can suggest feedback) without express, written consent from the student is forbidden under federal and university policy. With regard to other possible uses of AI for grading or providing feedback—i.e. ones that would not involve uploading student work—course heads and their teaching staff should discuss appropriate uses.
What else should we be thinking about?
The choices that you may make with regard to your approach to generative AI will have different repercussions for your course and your teaching team. Revising your assessments—whether to dissuade students from seeking assistance from AI, or to encourage them to reflect on what they can learn from it—may produce unanticipated, follow-on consequences. It may, for example, take longer to grade your new assignments, or be more complicated for students to complete and/or submit them. Here are some questions to consider:
-
If you opt to move more of your assessment to in-class modalities (e.g. oral or blue book exams), will you have the capacity to meet students’ accommodation needs? To grade oral presentations or handwritten exam books efficiently and equitably?
-
If you opt to encourage students to use AI tools in their assignments, their submissions may become longer or more difficult to assess—either because the marginal “cost” to students of generating additional material decreases, or because you, as the instructor, request that students incorporate more layers of meta analysis into the final product (e.g. generate a submission, and comment on what you learned in the process of generating the submission). You’ll likely want to consider how to manage workload issues for longer or more complicated assignment submissions.
-
One of the most effective ways to ensure that students are not over-relying on AI to complete their assignments is to ask them to produce reflective statements in which they unpack their process and describe how they’ve come to the conclusions at which they’ve arrived. Yet reflective statements can be challenging to grade, particularly if students are not taught ahead of time (e.g. through a rubric) how to make sure that their reflections are sufficiently analytical and evidence-based. You may wish to consider adding guidance to any assignments containing a reflective component that will ensure you are being equitable and transparent in your grading criteria.