AI in Academia: A Professor's Guide on When to Use It (and When Not To)
The Educator's Journey Through the AI Landscape: Opportunities and Cautions
The emergence of ChatGPT and the rapid adoption of Generative AI have introduced fresh challenges for educators. A key question arises: if machines are capable of completing student assignments, what is the impact on learning? I am open to the idea of integrating this technology into my classroom, but the path to doing so effectively remains unclear. Given the relative novelty of Generative AI, we lack established models for its educational applications. To begin developing these models, it's crucial that we, as educators, deepen our understanding of this technology—its capabilities, benefits, and limitations. Here is a summary of how I use it in my daily life.
Yes
Improving clarity in a sentence: First drafts are often cluttered and unclear, primarily because they are the stage where we're still shaping our thoughts. AI can expedite the writing process by structuring these rough ideas into a coherent form. It's important to note that AI assists with the structure, not the conceptualization – the idea still needs to come from your 🧠.
Prompt: “Improve this sentence, focus on clarity:”
Paraphrasing segments: After drafting a paragraph, I often use AI to refine its flow or to explore different phrasing options. This practice also aids in expanding my active vocabulary. As English is my second language, there are many words I understand but seldom use in my own writing or speech. You might wonder why I don't apply AI to entire articles. The reason is simple: when dealing with longer texts, tracking changes becomes challenging. AI might alter the intended meaning of a sentence, a risk that escalates with the length of the text.
I prefer Quillbot over ChatGPT for this function (no need to enter a prompt, just hit “paraphrase”).
Toning down emails: have you ever written an angry reply to an email? While I’ve never experienced this, I’ve heard it can be an issue. ChatGPT is an excellent Angry to Professional translator. It conveys the frustration at hand while staying professional (kind of like this TikToker).
Prompt: “Rephrase this email keeping a professional tone”
Creating titles: while coming up with innovative ideas is not the AI’s forte, it does a pretty good job creating titles for projects and articles.
Prompt: “Give me 5 creative/catchy/formal/informal titles for this article. Here is the abstract: [instert abstract]. It’s fun to discuss them with your collaborators.
Selling my work: I struggle to showcase my skills and achievements. In my annual progress report, I summarize all my work activities. ChatGPT has helped me explain how some of the activities contribute to my students, department, or college.
Prompt: “Write a paragraph summarizing my teaching activities and explain why they are important to students: [insert detailed bullet points with your teaching activities]”. Notice that here, it’s also crafting a paragraph. In report-style texts, it usually does a fine job.
Writing emails in a specific style: I had issues with some immigration lawyers and I wanted to make sure my emails were very formal and used legal jargon1.
Prompt: “Rewrite this email to my immigration lawyers using a legal style, emphasize the urgency of this matter”
Producing teaching materials: I teach interpreting (as in oral translation). And I need to create speeches for my students to interpret. Finding materials that are appropriate for their level can be challenging and writing a speech is time-consuming.
Prompt: “Write a speech on how cows contribute to climate change due to their methane emissions. Include a story to exemplify the main point. Write it with simple words, that could be easily understood by a 13-year-old. Include some figures”. The results are not perfect, and I wouldn’t do it to create an exam, but it does an ok-ish job.
Finding a word: one day, my boss popped into my office and asked me “how do you call this style of writing that gives a name and then qualifies it, like Achilles, he of the swift feet?”. I asked him if he had run it through My Friend (a.k.a. ChatGPT), and he replied “But how would I even ask it?”, to which I said: “exactly the same way you just asked me”. And that’s how we figured out that “Achilles, he of the swift feet” is an epithet.
Prompt: same as you would ask a person
Writing recommendation letters: certain parts of a recommendation letter are formulaic, exactly the type of writing where Our Friend shines. Two caveats: 1. if you want the letter to be decent, you’ll need to include very specific details, and 2. Do not include personal details, it’s still unclear what happens to the information you input there.
Prompt: “Write a recommendation letter for [program: ex. Med School] for an undergraduate student. Details on the student: - Took X course/s with me [when]. - Got X grade. - Projects they submitted that are relevant. - Personal qualities you want to highlight. Add a couple of lines on why this training is relevant for future studies in Med School, and two lines on how this student will diversify their student body (if they belong to a minority group)”. The more details, the better the letter. Bonus: if you can’t write a strong letter for this student, ask ChatGPT to write an email telling the student why you won’t be able to write the letter.
Coding: I only use R now and then to analyze my research data, which means that I forget everything and need to relearn it. ChatGPT helps A LOT. You can save yourself hours on substack.
Prompt: “Write me some R code to transform this dataset from horizontal to vertical”. “This code is giving me this error [insert error], how do I fix it?” Once, I even copied the description of a data transformation from a paper (the actual paragraph), pasted it in ChatGPT, and asked it to create the code to do the same with my data. The result wasn’t perfect, and I still had to modify it, but it got pretty close.
No
Writing grants: Even small ones. Useless. A grant draws on information from too many different sources. Providing a prompt that includes all the information you need to include is basically writing the grant. You’ll be able to apply some of what I described earlier (improving sentences, giving you title ideas, etc.), but ChatGPT won’t be able to get you millions in funding for your research (which is probably good news, your little brain is still a valuable resource). The same applies to articles.
Making decisions. I once bought two similar pairs of boots and asked the bot which one I should return (I included a list of pros and cons for each pair). It gave me a useless answer, just like a politician who doesn’t want to risk it. My sister-in-law gave me a much more useful answer.
Overall, texts produced by ChatGPT tend to be redundant and melodramatic, but this is an easy fix (delete the last paragraph, it’s usually a remix of things already mentioned, and delete overemotional words). I ran the first draft of this post through My Friend and he (it’s a he in my mind) told me that the topic was relevant, the anecdotes were engaging, and the structure was clear; but that I could improve in deepness (I agree, but I don’t think anyone wants to read 3000 words), get rid of the emojis and informal tone (nope 😒), and emphasize ethical concerns (fair point). Maybe next time. Bye 😗.
P.S.: If you’re wondering what I did with the boots, I kept both 😬 (thanks, Ángela).
Just to give you an example of how it transforms my sentence into legal: “Ensuring Formality in Legal Correspondence: Recently, I encountered some challenges with immigration lawyers, prompting the need for highly formal communication in my emails. It was crucial that these emails not only adhered to a formal tone but also incorporated appropriate legal terminology. To achieve this, I turned to AI assistance. It helped me ensure that my emails were precisely crafted, using the right legal jargon and formal style necessary for effective communication in a legal context. This tool proved invaluable in ensuring my correspondence met the required professional standards.” See how it’s a he? And he’s white, too.