Firas Moosvi, lecturer in the Department of Computer Science, explores the use of AI for generating his own class materials, and contemplates the ethics of such a choice, given that it draws from the hard work of other educators who have posted their materials on the Internet.
By Firas Moosvi
This website features stories and content submitted by individuals within the UBC community. We value the diversity of perspectives and experiences that these stories bring to our platform. Any views or opinions expressed in these user-generated stories are solely those of the individual authors and do not necessarily reflect the views or opinions of UBC as a whole. The presence of user-generated content on our website does not imply endorsement or agreement by UBC with any particular perspective.
Introduction
Over the last year, a lot has been written about generative AI, and how it is going to change the teaching and learning landscape. Much of the discourse has been limited to how students can use the tool with integrity and ethics. However, is it ethical for instructors to use the tool in course development and delivery? Unfortunately, this question remains mostly unresolved, largely because lawmakers have not yet – as of August 2023 – appropriately reconsidered (already archaic) copyright laws to incorporate outputs from generative AI.
With the absence of answers that are compliant with existing legal frameworks, the question becomes more philosophical – it is thus incumbent on instructors wanting to engage with the tool to do more soul-searching. This piece will guide you through my personal journey considering the ethics of using generative AI in my classes. For context, I am trained as a physicist, practice data science, and teach courses at the undergraduate level in computer science, data science, and physics.
ChatGPT Launches
On November 30, 2022, ChatGPT was released for public use. I found out about it soon after because I was active on Twitter a lot those days (thanks Elon, you killed a beautiful thing). I created an account, played around with it a bit, and realized it was a game-changer. Admittedly, I didn’t have enough imagination or foresight to even think about using it for teaching. I just thought it was amazing as a way to search for information, like Google 3.0 or something. My first few use cases were to answer coding-related questions that I would normally turn to StackOverflow for.
After playing around with it for a bit, I put it away for a while and let the information and prospects marinate in my mind. At the time, most of the deafening voices I heard about ChatGPT were panic and concern about how students would abuse the tool, and how much academic integrity violations would increase given this tool. Personally, I was less concerned about this, principally because of my teaching philosophy and practice of growth-based learning. My normal practice for assessments is to administer them digitally and allow students to complete them with open books, open notes, and access to the Internet. That said, ChatGPT was the first technology that truly made me reflect on the value of these types of assessments.
ChatGPT in my class
It was a bit unfortunate that such a powerful and disruptive tool was released midway through the semester, even though technology doesn’t usually follow the patterns of higher education terms. During that first term after ChatGPT was released, I decided not to prevent students from using the tool for labs and assignments (I didn’t make an announcement one way or another). Before the final exam started, I made it clear that students should not be using ChatGPT because we – as an institution – were still exploring the boundaries of the tool and what is allowed. To lean into the “still exploring the boundaries” portion, I included a question on my final exam that was inspired/generated by ChatGPT. Honestly, it was a bit uncomfortable and awkward for me to tell the students they couldn’t use a specific tool, only to turn around and use that exact same tool on the exam.
The genesis of this decision was interesting.
Using a question inspired/generated by ChatGPT on my final exam
A few days before the final exam, I was sitting in my home office, trying to set the final exam questions. I had used up all my inspiration and created all components of it, but was running out of ideas. On a lark, I decided to consult ChatGPT and started with this prompt:
"Come up with a coding exercise for first year undergraduate students about javascript."
I was blown away by the first answer I got! The question was incredibly detailed, clearly worded, and happened to be at exactly the level I was looking for. It even included some hints for the students, some test cases to explain how the function should behave, and – most importantly – contained a full solution! It seemed as if another instructor teaching a course very similar to mine had written a question and shared it with me – more on this later – for my final exam.
I was ecstatic. There was a moment where I considered if it was ethical for me to do this, but I decided it was fine to use it as long as I modelled appropriate use by adding a citation directly in my question:
Attribution: The author generated this question in collaboration with GPT-3, OpenAI's large-scale language-generation model. Upon generating draft language, the author reviewed, edited, and revised the language to their own liking and takes ultimate responsibility for the content of this question.
I cannot remember where exactly I got that prompt- I thought it was adapted from the ChatGPT FAQ, but oddly, that question and answer is no longer listed there! Though I had made the decision to use the question with minor adaptations, the more I thought about it, the more I felt guilt and regret.
Is it ethical for instructors to use ChatGPT for course development and delivery?
I felt guilt and regret because the corpus used to train the ChatGPT Large Language Model included the CommonCrawl dataset. This dataset is several petabytes large and contains a copy of all publicly accessible websites on the Internet. With this information, several questions arise:
- Did the instructor(s) whose questions the ChatGPT response was based on give consent to have their content included for the model to be trained on?
- Can the instructors whose intellectual property was used be credited or compensated in some way?
- Humans can adapt, remix, and rephrase content created by others (assuming it is an appropriate license). Assuming that the original source could be credited correctly, should the adaptations of a model like Chat GPT be considered similar to human adaptations and remixes?
- How is using a ChatGPT-generated question on an exam different from using questions from the back of a copyrighted publisher textbook?
Currently, the answers to questions one and two are definitively no.
The answer to question three is uncertain at the moment and requires further consideration.
For question four, on the surface, using a ChatGPT question is similar to taking a question from a textbook. However, Fair Use protections permit instructors to use copyrighted content from a publisher textbook. No such protections exist yet for content from generative AI.
Will it ever be ethical for instructors to use ChatGPT for course development and delivery?
Will Fair Use protections be granted to users eventually? Probably. Guidelines for the use of content from generative AI are being discussed at almost every legislative branch of the government. It’s a bit of a wild wild west out there at the moment but, in my opinion, using generative AI content is firmly in a grey area.
Is Fair Use the only aspect preventing me from using generative AI in my courses? No. I think there are other reasons as well. Most notably, it comes down to your philosophy and how you would answer the following question:
Are we doing a disservice to the spirit of open education by using AI-generated content with a dubious license when there are tonnes of content that exists with open and permissive licenses?
I say yes.
For now, anyway.
Post Script
I cannot imagine this simplified rationale and explanation will be palatable to others. There are dozens of factors to consider, and it is not always as easy as it sounds. The content that was scraped was publicly available, and if someone really wanted to prevent their content from being scraped, they could have set up their website to prevent scraping. Certainly, I don’t hold it against anyone who does choose to use generative AI in their courses; in fact, I did so myself and may do so again in the future!