How faculty and staff are using generative AI

In the arc of the going-on-three-years generative AI hype cycle, we appear to have entered thereality check phase. ChatGPT and similar platforms do indeed seem like they'll be sticking around, but forecasts that genAI will totally transform our lives and workplaces the way smartphones or the internet did now seem a bit, well, hype-y. Even as faculty continue to grapple with some of the problematic aspects of these tools, many people here at UM-Dearborn, particularly early adopters of the technology, are finding genAI platforms legitimately useful in their jobs, often with tasks that aren't that much fun to do.

In Information Technology Services, for example, the team routinely stares down mundane must-dos of office life, and they're increasingly turning to genAI applications to make those tasks less time-consuming. For example, ITS programmers are currently migrating the system they use to generate more than 300 reports that are used by various departments and units from one platform with a proprietary coding language to another platform that uses a different language. Given that genAI has proven itself to be pretty good with code, Applications Team Programmer and Analyst Thomas Stockwell decided to see if U-M Maizey, one of U-M's custom genAI tools, could help with the conversion. "It's not a magic wand and what you get may not be performance ready," he says. "But it's definitely something that can get you pointed in the right direction." Director of Applications Brian LaGoe says he recently talked with one of the programmers who's working full-time on this migration project, and he estimated that this genAI-based approach is saving one hour per report. "So you multiply that across 300-some reports and it's a real time saver," LaGoe says.

ITS is also using genAI in other ways. They've integrated a custom chatbot into the ITS staff's Slack, which enables service desk employees to do all kinds of things, from creating summaries of tickets to performing some initial troubleshooting based on ITS' knowledge base articles, which the bot has indexed. LaGoe says accessing information in this conversational way with the bot often helps service desk workers get the info they need more quickly compared with the standard "less-than-ideal" search tool that's built into the ticketing system's software. Two additional tools that are almost ready for deployment: An AI that automatically summarizes the notes that get created in the course of troubleshooting a ticket into a tidy package that could also feed the ITS knowledge base - a post-game paperwork task that is not a staff favorite. They're also working on a ticket triager that automatically routes incoming tickets to the proper team. LaGoe says they're also very close to launching a student-facing customer service chatbot that students could use as part of their OneStop experience.

Faculty are also finding uses for genAI. Accounting and Finance Lecturer Ahmet Tuncez is one of a handful of instructors who've activated the U-M Maizey "tutor-bot" in Canvas, the university's learning management system. Essentially, this custom bot can index everything an instructor has loaded into a course's Canvas page, like syllabi, modules, assignments, discussions, readings and lecture slides. Then, a student can interact with the bot in a conversational manner to get the information they need. Tuncez says the bot can handle super simple requests, like, "when is the midterm?" But students can also query things that are content specific, like explaining the concept of a hedge fund. He says one feature he really likes is that U-M Maizey provides a linked reference. "So it sort of functions like an index," Tuncez says. "It will give them links to the resources, such as a PowerPoint slide where that topic is discussed, so students can dig deeper if they need to." Tuncez also sees it as a win for him. Answering emails has become a bigger part of an instructor's daily routine, and he thinks U-M Maizey could handle some of the questions he gets. "I'm encouraging students to try it out. It can definitely tell them when the midterm is. And, honestly, I experimented with it a little bit and sometimes the answers are better than what I'm giving them," he says. "I'm half-joking. But it's really, really good. And it's 24/7." Given it takes 10 minutes to set up U-M Maizey in Canvas, Tuncez jokes that the payback period "is two emails." "If it saves you writing two emails, you've broken even. So why not?" he says.

Mathematics and Statistics Lecturer Hussein Nasralah, who has been experimenting with ChatGPT and other genAI apps for a couple years, has been using them a little differently. Interestingly, he doesn't use them for anything teaching-related. "I guess I still think my way is the best way. I'm pedagogically stubborn," he says. Instead, he's been using genAI for some of the more administrative tasks associated with his job. For example, when he learned about the new federal digital accessibility standards for online materials, he wanted to know what that was all about and what kind of changes he might have to make to his math courses. So he asked ChatGPT to create a quick start guide that explained the rationale behind the new standards and what he'd have to do to bring his course into compliance. He said he found what ChatGPT produced so helpful, he actually shared it with a few colleagues, who agreed. After that, he actually began using ChatGPT to convert his course materials into accessible formats. "I have this PDF article that I have my students read early in the course about how the brain is like a muscle, but this PDF is ancient and it's not accessible," Nasralah says. "So I asked ChatGPT to help me turn this into Canvas-friendly HTML code and make it accessible for the new federal guidelines. I took screenshots of the pictures and told it to leave placeholders in the HTML so I could add them later. I had it generate the alternative text for the images. And then it gave me the HTML." Nasralah says he did have to spot check the text later, because ChatGPT couldn't resist making some subtle wording changes. But the output gave him HTML that was more or less cut-and-paste ready for Canvas. It saved him hours of tedious work.

Nasralah also used Google's genAI tool, Gemini, to create online math worksheets for his students. Historically, he says math faculty have used a platform called WeBWorK to create digital math homework, but it's an older platform that runs on a computer language called Perl. "I definitely don't know Perl, so that's always held me back from creating my own WeBWorK assignments from scratch," Nasralah says. He asked Gemini to take a PDF of a basic algebra skills assignment for his Calculus I course and turn it into Perl code, so he could cut and paste it into WeBWorK. That worked too.

The faculty and staff who are finding genAI tools helpful tend to be early adopters who've been playing around with the technology for a few years. In Nasralah's case, he's used them for all kinds of things in his professional and personal life. Right now, for example, he's also using ChatGPT to build a meeting scheduling website that has all the features of his favorite scheduling website plus a few he wished it had. He's been able to do this, without knowing HTML, because in the course of using ChatGPT for hundreds of hours, he's gradually developed instincts about how it works and which prompts will give him the best results. So he understands why faculty who might not have the same natural enthusiasm for new technologies and haven't developed the same literacy haven't found genAI that helpful yet.

LaGoe, who also falls into that bucket of an early-adopter technophile, says he totally understands that sentiment too. But he feels comfortable recommending at least one genAI tool that he thinks most people could be using: Zoom's AI Companion. "I know I don't take super great meeting notes, and it does a pretty amazing job of summarizing conversations and even picking out action items from meetings," LaGoe says. "Recently, we even had an academic advisor reach out, and they shared that they were having their students turn that feature on for their advising sessions. It did a great job of summarizing what they talked about, which gave the student a record of everything they needed to remember from that conversation."

###

Are you a faculty member who has questions about using generative AI in your course? If so, you can always reach out to the staff from the Hub for Teaching and Learning Resources. Story by Lou Blouin

Share

• facebook
• twitter
• linkedin