Guidance for Using Generative AI in Coursework

There has been a great deal of press and confusion about terminology related to GAI tools, how they work, and how we should think about the opportunities and risks related to them. For a curated summary see “The Benefits and Limitations of Generative AI: Harvard Experts Answer Your Questions” and links from there to other articles such as “The Future of Generative AI.”

All instructors are expected to include a GAI policy in their course syllabi and post it on their Canvas course sites.

Most courses already specify policies for whether and when students may make use of work that is not their own, such as secondary sources or ideas shared in a study group. Instructors should adapt such policies to include work done by GAI. For some courses, it will make sense to permit all uses of GAI; for others, to prohibit all; and for still others, to permit some uses and prohibit others. Policies should specifically define the use and misuse of GAI for that course to ensure that expectations are clear to students, and should be aligned with the Harvard Chan School’s academic integrity policies.

Instructors are encouraged to discuss their policies and the rationale for them at the beginning of their course and/or as relates to specific assignments.

Unless otherwise specified by an instructor, it is a violation of the Harvard Chan School’s academic integrity policies for students to use GAI to create all or part of an assignment for a course and submit that work as their own. This rule parallels other rules. Students may not ask another person to complete their assignments for their courses. Students may not copy something someone else has created, or re-write something someone else has written, without proper attribution.

In general, the School sees permissible uses of GAI in coursework to include seeking clarification on concepts, brainstorming ideas, or generating scenarios that help to contextualize what a student is learning. This may include, for example, using AI-powered web searches, having “conversations” with tools like ChatGPT to help explore ideas, refine thinking, identify examples, and better understand course material. It is also considered acceptable to use GAI to draft emails to instructors, students, and others in the community that are not being submitted as coursework.

If students have any doubt about whether a specific use of GAI is permitted for an assignment or course, they are responsible for discussing it with their instructor prior to using it.

The policies below include three short examples that are adapted versions of those developed by the Harvard Office of Undergraduate Education and one longer example from a Harvard Chan faculty member. Instructors are free to adapt the language of these policies for the needs of their courses.

Maximally restrictive draft policy

We expect that all work students submit for this course will be their own. In instances when collaborative/group work is assigned, we expect for the assignment to list all team members who participated. We specifically forbid the use of generative artificial intelligence (GAI) tools at all stages of the work process, including preliminary ones. Violations of this policy will be considered academic misconduct. We draw your attention to the fact that different classes at Harvard Chan School (and other classes across Harvard) could implement different GAI policies, and it is the student’s responsibility to conform to expectations for each course.

Fully encouraging draft policy

This course encourages students to explore the use of generative artificial intelligence (GAI) tools for all assignments and assessments. Any such use must be appropriately acknowledged and cited. It is each student’s responsibility to assess the validity and applicability of any GAI output that is submitted; if you choose to use the tools, you bear the final responsibility for the work product, including appropriate citations. Violations of this policy will be considered academic misconduct. We draw your attention to the fact that different classes at Harvard Chan School (and other classes across Harvard) could implement different GAI policies, and it is the student’s responsibility to conform to expectations for each course.

Mixed draft policy

Certain assignments in this course will permit or even encourage the use of generative artificial intelligence (GAI) tools. However, please be aware that the default expectation is that using GAI is NOT permitted unless the instructor explicitly states that it can or should be used for a particular purpose. Any use of GAI must be appropriately acknowledged and cited. It is each student’s responsibility to assess the validity and applicability of any GAI output that is submitted; if you choose to use the tools, you bear the final responsibility for the work product, including appropriate citations. Violations of this policy will be considered academic misconduct. We draw your attention to the fact that different classes at Harvard Chan School (and other classes across Harvard) could implement different GAI policies, and it is the student’s responsibility to conform to expectations for each course.

Harvard Chan example

Two of our Faculty Instructional Coaches, Garrett Fitzmaurice and Heather Baer, worked collaboratively to develop GAI policies for their courses; here is one of them.

EPI 522 Analytic Methods for Epidemiology

Collaboration
Because many students learn best when working in a group setting, we encourage collaborative learning in this course. You are allowed to discuss the weekly homework assignments with other students, but you must submit your own answers in Canvas. Also, please be aware that we are using the “shuffle answers” option in Canvas, which will randomize the order of the question’s answer choices. This means that students will get the same question, but their answer choices will not be in the same order.

For the two exams, you must complete them individually; you may not discuss them with other students or with people outside of the class (although you may ask the instructors if you need clarification on any questions). All of your work on the exams must be your own, in your own words, reflecting your understanding of the material. Solutions prepared “in committee” or by copying, paraphrasing, or summarizing someone else’s work are not acceptable. All computer output you submit must come from work that you have done yourself; handing in output from someone else’s computer session is not acceptable.

Use of Generative Artificial Intelligence (GAI) Tools
When trying to develop proficiency with statistical software, the use of generative artificial intelligence (GAI) tools (e.g., ChatGPT) may be helpful for creating relevant sample code that may only require some relatively minor customization or modifications to produce the required analysis. This type of use of GAI tools, for the creation of computer code for statistical analyses, is permitted and may supplement the explanations of relevant Stata code provided in the course. However, be careful not to over-rely on computer code produced by GAI tools; it is also important that you understand the different options and variations in syntax that may be required for your particular analysis. Note that a major emphasis of the course is on the interpretation of the output from statistical software. When completing all assignments (including the two exams and all assignments for the group project), these written interpretations must be your own, in your own words, reflecting your understanding and reasoning; using text written by a GAI tool as one’s own in the interpretation of your output is not permitted. With that said, the use of GAI tools to check and improve the grammar and sentence structure of your written work is acceptable; this use of GAI tools may be particularly beneficial for students for whom English is a second language. Finally, any use of GAI when completing assignments must be appropriately acknowledged and cited; appropriating a GAI tool’s outputs without giving credit amounts to plagiarism and will be considered academic misconduct. Guides on citations of the following styles are available:

• • • • MLA: How do I cite generative AI in MLA style?
      • APA
        • ChatGPT & Other AI Generative Tools
        • How to cite ChatGPT
      • Chicago Manual Style: Citation, Documentation of Sources

Please see the Student Handbook for additional policies related to academic integrity and disciplinary actions.

1. Seek to understand how GAI tools work, including their strengths and weaknesses, to optimize their value for student learning.
2. Have a course policy. Include it on your syllabus and course Canvas site. Be explicit and transparent about your grading policies and how use of GAI tools influences those policies. For example, if you permit use of GAI, do you plan to treat work by students who declare no use of GAI tools as the baseline for grading? Make clear if you plan to give a zero on assignments where students simply reproduce the output from AI tools or if you plan to impose a penalty for regurgitation of GAI outputs with limited evidence that students have meaningfully engaged with the content. Make explicit whether you would give extra weight to students who use GAI with creativity and critical nuance.
3. Consider how you might train students to use GAI tools effectively, in recognition of the fact that our graduates will likely be called upon to use these tools in their workplaces. For instance, you might ask students to use a GAI tool to generate a policy argument or a grant application, then use their own critical thinking skills to critique, revise, and improve on the GAI’s output.

1. Know the GAI policies of your course. Do not use GAI tools if an instructor explicitly prohibits it on a given assignment or exam or as a blanket policy in a course.
2. If you use a GAI tool to complete a course assignment, including use to improve the readability or language of your submission, you must acknowledge and document their use. In your assignment submission, include a statement explaining what tool(s) you used, prompts you provided (if applicable), and how you integrated the output into your work. Appropriating a GAI tool’s outputs without giving credit amounts to plagiarism and will be considered academic misconduct.
3. Recognize that GAI tools are prone to producing misleading outputs. The information provided by GAI tools is generated from unverified crowd-sourced information. Large language models can make false claims or “hallucinations” and will regenerate any biases in the corpus of texts on which they are trained. You therefore should not trust the information as if it were equivalent to published research. You are ultimately responsible for the accuracy of the work you submit.

The use of GAI has implications for the protection of your own intellectual property. For example, if you upload your own original content to a generative AI tool, that content may become part of the tool’s models, which others may encounter and use. Conversely, if you use generative AI to develop your own original work, it may unexpectedly include others’ copyrighted material.

Certain uses of GAI infringe on copyright laws applicable to U.S. universities or contravene existing expectations for student conduct in courses. For example, students may not post, publish, sell, or otherwise publicly distribute course materials without the written permission of the course instructor including lecture slides, video, and audio recordings, assignments, problem sets, examinations, other students’ work, and answer keys. Students may not make recordings of course material for their own use without written permission of the instructor.

Therefore, uploading substantial course content – including text, video, images, readings, discussion-board pages, assignments, or audio recordings – to a GAI system is not permitted. Exercise caution when submitting short excerpts from course materials when interacting with GAI.

Students are not permitted to record course meetings, with or without GAI tools, without permission. If students require or prefer that course meetings be recorded, they need to discuss this request with their instructors; if GAI technology is part of an assistive technology solution to enable students to participate fully in a course, this usage must be coordinated and approved through the Office of Student Affairs.

Credit use of GAI tools whenever used. During this time of rapidly evolving capacities of GAI tools, it is better to give credit broadly and inclusively, e.g., even if used only to generate ideas rather than usable text or illustrations. When using AI tools on assignments, students (and instructors) might consider the value of including (or assigning) an appendix that shows how and why GAI was used.

The guides on citations of the following styles are available.

• MLA: How do I cite generative AI in MLA style?
• APA
  • ChatGPT & Other AI Generative Tools
  • How to cite ChatGPT
• Chicago Manual Style: Citation, Documentation of Sources

See also:

• STEM LibGuide Resources: Citing ChatGPT, Library, University of California, Santa Barbara
• ChatGPT and other generative AI tools, Library, University of Queensland
• ChatGPT and Generative Artificial Intelligence (AI): AI-generated content and citation, Library, University of Waterloo

Harvard University Information Technology has compiled a list of available GAI tools.

In some cases, GAI tools may help students for whom English is not their primary language or students who could particularly benefit from templates, or from formatting and grammatical aides.

It is expected that subscription-based AI tools could surpass free tools in capacity, raising concerns about equitable access to this technology.

There are a variety of tools that claim various degrees of success in finding instances when GAI was used, but they are flawed and there are many examples of false positives and false negatives using these detection tools.

It would be inadvisable to rely on automated methods for GAI detection. Harvard Chan does not plan to provide/license such a tool for use in courses.