Principles for Students Using Generative AI
This is a list of sample principles that faculty may want to use in their classrooms. It is meant to be adapted to the needs and learning outcomes of the course it is used in.
In what follows, every
question
is meant to be a required question with only one choice: a verbal “yes.”
In [course name] this semester, we understand that many of you will want to use ChatGPT or another similar tool in your work. We want to make sure you are aware of your responsibilities with regard to using these tools, so we ask you to sign this agreement. However, you are not required to use ChatGPT at all! We require you to sign it preemptively so that, if you decide to use it, you will not have to go through this process then.
Do you understand that you are not at all required to use ChatGPT?
The content that you enter into ChatGPT will be retained into OpenAI’s training data.
Do you understand that ChatGPT will keep your interactions with the system?
You should not assume this agreement extends to any other courses. Other courses may have very different policies on the use of such tools.
Do you understand that this agreement is limited to this course?
Chapman University’s Academic Integrity Policy states, “Students are responsible for doing their own work, and academic dishonesty of any kind will be subject to sanction by the instructor/administrator and referral to the University's Academic Integrity Committee, which may impose additional sanctions up to and including expulsion.”
Do you understand you should add comments or other notations indicating how you used ChatGPT?
ChatGPT is a very experimental technology and it does not “understand” any queries we write. Therefore, its answers may be grossly wrong or, worse, subtly wrong. You take all responsibility for work you turn in, whether AI-generated or not. “The AI wrote my homework” is not an acceptable excuse for mistakes. ChatGPT is, at best, only an aid, and at worst, could even be an obstacle.
Do you understand that you take final responsibility for all work you turn in?
It is tempting to ask all kinds of random course-related questions to ChatGPT. However, many questions are much better addressed in class instead. The answers you get in class may take a little longer but will be authoritative. The answers from ChatGPT will be immediate and on occassion possibly very wrong. Use the right tool for the right job.
Do you understand that for many course-related queries, ChatGPT is a poor source of information and may even mislead you?
Remember that OpenAI intends to use the queries you enter into ChatGPT as training data. Therefore, you should never share any personal information in these queries.
Do you understand that OpenAI will read your queries, so you should not share personal information?
ChatGPT may become overloaded or otherwise unavailable. We will not give extensions or other accommodations if this happens.
Do you understand that there are no promises or obligations of ChatGPT’s availability, and that its unavailability (collectively or individually) will not be a source of accommodation?
ChatGPT is trained on large textual datasets produced using the open internet as a source. As a microcosm of the internet, this data may include biased and offensive language and assumptions. This data may also include disturbing (e.g., sexist, racist, violent) material. In an effort to “clean” the data that ChatGPT was trained upon, OpenAI outsourced this work to Kenya and paid the workers approximately $2.00 an hour for their labor, work which has been shown to cause PTSD.
Do you understand that the creation of tools such as ChatGPT depends upon hidden labor performed in situations where protections for workers may be inadequate?
Finally, the use of powerful tools remotely through the Web can easily hide the environmental costs of using them, whether search engines, ChatGPT, or other services. Therefore, we ask you to think about the environmental consequences of using such tools.
Do you understand that, in the aggregate, there can be significant environmental impacts from using tools like ChatGPT?
[This version of this document was created by Doug Dechow and Taylor Greene. If you use this document, we ask that you contact us so we may track its usage.
The original document was written by Shriram Krishnamurthi, with valuable feedback Ben Greenman james_tompkin@brown.edu Mark_Dieterich@brown.edu michael_littman@brown.edu timothy_nelson@brown.edu.
Feel free to copy and modify it as you wish. ]