ChatGPT AI impact on Teaching and Learning

The original communication from the Office of the Provost was sent out on Jan 23, 2023 to guide instructors at UConn.

Addressing AI ChatGPT

Many faculty have reached out seeking guidance surrounding possible academic integrity issues triggered by the recent release of ChatGPT3 from OpenAI[1].

For those of you who are unfamiliar, Chat GPT3 is a Large Language Model tool developed by OpenAI. The tool was released in November 2022, is powered by large amounts of data (~175 billion parameters), and relies on deep neural networks to predict and generate text in response to user prompts. There is much debate about the quality of the text generated but there is both interest and concern related to AI-generated text.  Based on conversations to date, our faculty are simultaneously interested in learning how ChatGPT3 and similar chat bots might transform teaching, learning, and assessment in innovative ways, and concerned about students use of Chat GPT3 to answer test and exam questions and generate content for written papers and assignments. Below, we lay out a number of recommendations for our faculty.

  1. Faculty members might take time to experiment with ChatGPT[1] for themselves and discover the capabilities and limitations for their particular discipline and context. One way to become more familiar with ChatGPT is to submit your own writing prompts or assignment questions and examine the type and quality of text generated. Making multiple attempts can illuminate the range of text generated to the same prompt. In addition, don’t be surprised if you get a message ‘ChatGPT is at capacity right now’.
  1. Faculty are encouraged to engage their colleagues and students in conversation about ChatGPT. This will help us to crowdsource both pedagogical opportunities and assessment concerns.  Some faculty plan to have students use ChatGPT to generate essays and then have students critique and improve those essays[2]. Other faculty are busy revisiting their assessment strategy and assignment particulars. Still others told us they plan to ask students to provide writing samples at the beginning of the semester.  Another indicated they planned to replace some writing assignments with oral presentations with Q&A, or contingently reserve the right to do so. 
  1. Faculty are encouraged to amend their syllabus[3] to add a statement that explicitly addresses student use of ChatGPT or similar tools use that is consistent with their teaching and learning philosophy. Faculty should be transparent about what students are permitted to use and under what conditions. Faculty might consider requiring students to include brief acknowledgement statements with assignments, ones in which they share (and perhaps reflect on) which tools they used in the process of completing their work. Two divergent examples are provided below and additional samples will be accessible on the CETL syllabus template.
  1. Every faculty member should take advantage of this opportunity to reflect on their course and lesson learning outcomes and (re)consider the extent to which their assessment strategy is aligned to measure those learning outcomes, and to identify adjustments needed for existing assessments and opportunities for creating alternative authentic assessments.

We believe maintaining a balanced and realistic perspective of the impacts of AI is more productive and appropriate than a narrow focus on surveillance and detection. While we believe that the vast majority of Uconn students are committed to the highest ideals of academic integrity, we also know that some students may be tempted to use ChatGPT3 and other tools in ways not permitted. For this reason, we wanted to let you know that the University administration is exploring various AI detection software that faculty could use to identify possible AI generated content. Until then, faculty may choose to use open source experimental detection applications such as GPTZero and AI Content Detector, being mindful that interpretation of results should be tempered by the fact that neither provide definitive evidence.

Finally, we encourage all faculty to keep abreast of messaging and developments related to ChatGPT this semester. Check for more workshops and panel discussions on the topic forthcoming in the coming weeks and months.

[1] https://openai.com

[2] https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4300783; https://medium.com/@rwatkins_7167/updating-your-course-syllabus-for-chatgpt-965f4b57b003

[3] See examples below or visit https://cetl.uconn.edu/resources/design-your-course/creating-your-syllabus/

Sample Syllabus Language

Addressing student use of AI ChatGPT

Sample # 1 (Permitting use of AI ChatGPT)


Academic Integrity
In this course we’ll conduct ourselves as a community of scholars and writers, recognizing that academic study is both an intellectual and ethical enterprise. Please build on the ideas and texts of others–that’s a vital part of academic life. You may certainly discuss readings and assignments outside of class, study in groups, share drafts with classmates or friends, and go to the Writing Center with your drafts.

When you use or borrow or closely imitate another’s ideas or language–or even syntax–you must formally acknowledge that debt by signaling it with a standard form of academic citation. This means documenting not just direct quotations but also paraphrases and summaries. In less formal or creative genres, you may show your debt to a source (or classmate!) with a signal phrase (“According to Jose Calabra….”) or acknowledgement statement (“In this essay I drew inspiration from…I got the_____ idea  from Kayla during peer review.”). If you have any questions about when and how to credit the work of others, please come talk to me.

You are welcome to use AI writing tools such as ChatGPT on most assignments (I’ll alert you when you can’t) but whenever you use them, you must include an acknowledgement statement that briefly shares that and how you used them. For example, “I used ChatGPT when I was struck at the start and retained substantial parts of what it produced, including X and Y ideas and most of the wording in paragraphs 3 and 4” or “After I wrote my first 2 paragraphs, I used GPT-3 playground to extend the text for another 200 words but then edited…” Please also note that all large language models still tend to make up incorrect facts and fake citations. You will be responsible for any inaccurate, biased, offensive, or otherwise unethical content you submit, regardless of whether it originally comes from you or an AI tool (these last 2 sentences adapted from the course policies of Ryan S. Baker, University of Pennsylvania).

If you engage in intentional academic dishonesty–whether plagiarizing or submitting the work of others or copying from others on a test or failing to acknowledge use of AI or other tools–you will fail not only that assignment but the course.

Sample # 2 (Prohibiting use of ChatGPT)

All students are expected to act in accordance with the Guidelines for Academic Integrity at the University of Connecticut. If you have questions about academic integrity or intellectual property, you should consult with me or consult UConn’s guidelines for academic integrity. Posting course material on student tutoring and course sharing websites (e.g. Chegg, Course Hero) may be a violation of my copyright and intellectual property and a violation of academic integrity. Many of you may also be aware of the recent release of ChatGPT3, a Large Language artificial intelligence (AI) model that has the capacity to quickly produce text on a range of topics.  ChatGPT3 aggregates the ideas and insights of many researchers without giving them credit.  Submitting ChatGPT-generated text as your own work would be an act of plagiarism insofar as it would involve passing off the work of others as your own. For these reasons, you are not allowed to use this ChatGPT or other similar tools.to produce essays or other academic work for this class, unless otherwise explicitly permitted to do so.  You should also know that the university has AI detection software that distinguishes between AI generated content and human generated content.

For additional samples, see Lance Eaton's collection of syllabus policies.