Gen-AI basics and tips for teachers in higher Ed, English in particular

What to know

  • Gen-AI is "generative artificial intelligence," computer programs that generate text, images, and/or code. When your phone suggests what words you're about to type, it's using Gen-AI.
  • There are many large-scale programs now available to us and to students: those that write/inform as well as revise/correct (ChatGPT, Gemini, Lex, Bing...) and those that mainly revise and correct (Grammarly, Quillbot...). Many have free basic services and monthly subscriptions for better services.
  • The programs get better constantly and have access to more information as they upgrade. It is no longer true that ChatGPT et al. will only know things up to 2021, and they're now becoming able to access work behind paywalls.
  • Gen-AI programs work from "prompts," which are questions, requests, and clarifications that you give them. How you word the prompts affects output (like using different search terms in a database).
  • Gen-AI that writes and informs also "hallucinates": this is the term for what happens when a program makes up something plausible to answer your prompt. For example, if you ask ChatGPT to include quotations from peer-reviewed articles in an essay, it will find the name of a critic who has published in that area and the name of a journal that publishes in that area, and it will fabricate a title, page numbers, and a quotation. 
  • Whatever you give a Gen-AI becomes data that its company owns and monetizes.
  • Yes, OpenAI (the company that owns ChatGPT) used low-paid Kenyan workers to watch violence and pornography to teach ChatGPT what was toxic content.
  • Yes, Gen-AI uses a lot of resources and does environmental damage through its storage facilities…and so does everything cloud-based, every show you stream, etc.
  • Yes, Gen-AI exhibits lots of prejudices, blind spots, and general nastiness: it is trained on the internet!

What not to do

  • Don't forbid students from using all Gen-AI programs: you can't police it (detection software is unreliable), you probably don't want to police it, and our students need to gain skills at using Gen-AI in most of the jobs/careers they'll have with a degree in English.
  • Don't just avoid mentioning it: students are worried that Gen-AI will affect their job prospects, so they want information; also, it's very tempting to mis-use Gen-AI in desperation, so you need to make clear what's acceptable and what's not acceptable.
  • Don't put your own intellectual property into a Gen-AI. 
  • Don't put your students' intellectual property into a Gen-AI.
  • Don't put a living author's intellectual property into a Gen-AI.

What to do

  • Use some of the free programs yourself, experiment, learn to design prompts, and put your current assignments into a writer such as ChatGPT to see what happens.
  • Have a detailed and clear course policy, including what aspects of Gen-AI students can use and why, but that they should not use it to generate written assignments de novo. Gen-AI is good at "fixing" grammar to an outdated colonial formal English, which is beneficial for many students. However, you should warn students about the hallucinations and let them know that if they put their own work into a Gen-AI they compromise their intellectual property.
  • Redesign any assignments that the programs can do passably well. This could involve 1) making them more creative and personal, because Gen-AI is not good at that yet, 2) posting students' work in Canvas for everyone in the class to see/review, because peer pressure should not be underestimated, and/or 3) working on assignments in person in class, and in different stages.
  • Demonstrate in class what Gen-AI can and can't do well. The English dept here at SFU has a blanket PIA (privacy impact assessment) for demonstrating ChatGPT, so long as you don't put any student's ideas or personal information into the program during the demo. If you want to demonstrate another program that is not approved to be embedded in Canvas, you'll need another PIA. If you require students to put their own work into ChatGPT, you’ll need a higher-level PIA.
  • If you design an assignment in which you require students to use ChatGPT or another Gen-AI, keep in mind that some students will be disadvantaged if they cannot afford the deluxe subscription versions that others in the class have. And warn all students that they will be giving their intellectual property to the company without credit or compensation.
  • If you suspect a student has used Gen-AI to generate an assignment de novo, treat it as you would a case in which you suspect a student has asked or hired another human to write their work: in the interview, ask them to explain what they were intending/meaning in different parts of the work, ask to see rough drafts, ask to see a history if they used google docs, etc. If the student says they used Gen-AI to revise, correct, give feedback, etc., but not create the ideas, ask to see the prompts and their results.

There’s a ton of info available online about Gen-AI, and the field is changing quickly. If you want a decent starting place, try this series of short videos from Wharton.