So, you're wondering what is an appropriate use of AI in academia, eh?

NOTE: This is part of a series of posts that I've written for students. Here are some others:

• "So, you want me to be your graduate supervisor, eh?"
• "So you want me to write you a grad school reference letter, eh?"
• "So you're putting together a Research Outline, eh? Here's a flowchart that may help!"

Here’s my personal policy on using Artificial Intelligence in academia: Asking an AI Chat Bot to write your assignments is a form of academic dishonesty (just as it is unacceptable to search online for existing text and submitting that work as your own). There are, however, ways to use AI appropriately as a support for research and writing (just as it is appropriate to ask colleagues or others to support your academic work in one way or another). We will have a discussion in class about this to make sure we are all on the same page about what constitutes a legitimate use of AI in your assignments. 

As an example of appropriate uses of AI from my own experience: I sometimes ask my AI to help suggest clearer phrasing when I’m struggling with a long and difficult sentence. Or, I’ve uploaded text I’ve written and asked it to critique the argument and outline areas of weakness. I also have used AI to help manage a data file in Excel (creating new tabs, formulas, and charts to cut down on the amount of manual labor required to do this myself), and to look something up on the internet (for instance, “what did X person have to say about Y? Can you provide a link to the source?”. I consider all of these appropriate uses of AI. However, in each instance I want to emphasize three things: 

First, AI isn’t necessarily better. It is important to me personally that I don’t come to see AI as superior to human work – after all, it is fundamentally based on human language, and built by humans. Also, I do not want to rely on AI as a crutch, which I feel could dull my critical thinking capacities. I want to use it as a tool to support further critical thinking skills. So, I recommend always asking oneself “do I really need to ask this of the AI, or can I do it better if I just put in the effort?” Many times I have found using AI has actually taken up more time than if I had just worked out a problem myself. 

Second, it is important to understand that AI hallucinates and sometimes makes things up. It is based on existing language out there in the world, but it can be duped, frequently misunderstands social cues or human context, and tends to make up answers rather than admit that it does not know. As an example, I once asked an AI Chat Bot to create a list of the 10 most influential books on a given topic. It came back with a list, but some of the entries were of questionable fit (not quite related to the topic I had asked), and a few of the entries were completely fabricated! They were real-sounding books by real-sounding experts, but they totally did NOT exist in the real world! All this to say, it is absolutely crucial to fact check and verify any substantive work that you ask AI to do, especially if this involves internet research. As scholars we have a duty to verify work produced by AI – especially if our work ends up online – to make sure that future research is not building upon faulty and fabricated data. 

Finally, I would like students to add an AI usage disclosure statement to any assignments for which AI was used. This should appear at the end of your assignment as a statement indicating how you used AI in your research and writing process. If you should find yourself asking your AI Chat Bot to write out an AI usage disclosure statement for you, I recommend that you stop, step away from your computer, and go for a walk in nature, and contemplate what the world might be like if humans no longer feel confident writing anything without first asking an AI to do the work for them.

AI Usage Disclosure Statement: I uploaded the policy above to my AI Chat Bot and asked it to provide suggestions for improvement. While it did come back with a list of ideas, I decided none of the suggested additions it offered were really necessary, so the original human-generated text has remained unaltered.