PITA tasks, low hanging fruit and AI
75% > 0%
Over the past few weeks, I've talked with several groups of educators about AI. One of my main messages to these groups is to start with the proverbial low-hanging fruit ... applications AI that are relatively easy to implement but that provide value. In this case "provide value" usually means one of two things: doing something that you couldn't easily do without AI, or (more likely) doing something that you hate doing. What I call PITA (pain in the a$$) tasks. For example, rubrics are a pain to create but sometimes necessary. So is checking instructions or other documents for small errors. I don't think I've ever created a syllabus without a single error; maybe AI can help.
Fortunately, AI is pretty good at many PITA tasks. To get started with AI, I recommend focusing on your PITA tasks starting with the one that's easiest to implement. For me, this would probably be writing exam questions. They're important. They're part of the job. But creating them is boring drudgery. So, I've offloaded this to AI. Now, I just make a pdf of my lecture slides, upload the file and ask AI to crank out a bunch of questions with varying levels of difficulty. Usually, I ask for more questions than I need since AI will produce some that I don't like. Creating the exam is then a matter of choosing the questions I like, making a few edits, and putting them all in a Word file. Usually, I generate questions one chapter/lecture at a time to make sure the exam is balanced. Also, I may add a few questions of my own. In relatively little time, I have a pretty good exam. A PITA task is now less of a pain. For me, exam writing is a sweet spot where PITA and AI come together.
There's are a couple of important messages here. First, this application is pretty easy to do with AI. You don't need to be a "prompt engineer." You just need to know how to give some simple instructions. My prompt for creating exam questions basically matches what I would tell a teaching assistant.
Create X number of short-answer (or whatever) questions based on this material. Make some of them easy, some hard, and some in-between.
That's pretty much it--nothing complicated, nothing that requires deep AI skills. Just some simple instructions. AI can help with many other PITA tasks just as easily. A colleague has Microsoft Copilot crank out routine emails and replies. Of course, he edits them (and discloses the AI use).
The second important, but kind of hidden message is that your use of AI will expand naturally. Over time, you'll learn more about AI and will start to see new opportunities to use it effectively.
Try this little exercise over the next couple of weeks. Every time you find yourself dreading a task, write down the task. Once you have a bit of a list, go through the tasks and see if AI can help with any of them. If so, pick one and start using AI. Then, try with the next task and so on. Over time, you'll reduce the time you spend on PITA tasks. The best task to pick is one that is at the intersection of PITA-ness and ease of AI application. In other words, pick one that's a major pain but is easy to do with AI. You'll need to know a little about AI to know which tasks to pick, but you won't need to be an expert. If you haven't already gotten started with AI, check out my guide, which is available to subscribers. Or if you really don't want to subscribe, email me at craig@AIGoesToCollege.com and I'll send it to you. It's free (really).
Here's a pro tip: DO NOT try to get AI to do 100% of the task. This is where many people go wrong. They expect (or hope) that AI can do the entire task. That might be possible, but it's usually not easy. I'm fine with removing 75% of an onerous task. (Onerous is such a great word.) 25% PITA is a lot better than 100% PITA. Often, getting AI to do a good portion of something is much, much easier than getting it to do 100% of something.
What are you waiting for? Find your low hanging fruit and get started. Before you know it, the PITA factor in your life will be greatly reduced. And that's a good thing.