The arrival of ChatGPT has put colleges in a panic, fearing students won’t learn to research and write. Educators now understand that banning AI programs from the classroom isn’t the answer. One teacher at Wright State is already preparing students for a future of writing with artificial intelligence programs. David Seitz has the story.
Alan Knowles started researching the future impact of Artificial Intelligence on writing five years ago. At first, he worried about the potential for student cheating and the spread of disinformation. Since then, he has brought AI into his technical writing classes at Wright State University.
Knowles said, “Once I started seeing how students were using these tools and what they were able to do with them, my panic really subsided.”
Students learned it took effort to get good results. Lauren Eiffert is a student in technical writing. For an assignment, she wrote a user’s manual for the espresso machine at her work place. ChatGPT wasn’t very helpful. Eifert said ChatGPT “reacted as if I was talking about a Mr. Coffee coffee pot instead of a five-thousand-dollar espresso machine that’s so intricate and complex.”
Knowles says students still require expert knowledge of a subject to achieve strong writing with AI tools. “If you are a history professor, you are still going to be able to use ChatGPT to write a better essay than your history students because you know more about the content. So you will be able to prompt the AI better than your students will at this point. You will also be able to fact check what it generates better than your students, and that takes a lot of work.”
ChatGPT and other AI writing tools are a large language model, an enormous database of previous internet text. Large language models don’t think in words. We give it words, and it processes them as tokens. The AI then transforms them back into our language. “They are imitating our language,” commented Knowles. “It’s basically a statistical model of language. For every single word, or token, it’s going to have a probability percentage of what comes next.”
The AI needs careful prompting, so it doesn’t generate false information. Knowles helped his wife use AI to write a fund-raising letter to buy new equipment for a hospital. First, they gave the AI a bulleted list of information.
“It stuck to all the information that we gave it in our very detailed outline,” remarked Knowles. “So in this sense, we weren’t using the AI to give us information or to make up facts for us. We’re using it specifically as a writing assistant to do some of the labor for us.”
Knowles says that AI can help small non-profit organizations that can’t afford professional writers. He is also teaching technical writing students how to train the AI. This is a new skill called “prompt engineering.” Lauren Eifert worked with a team to generate social media posts to promote Wright State. The team learned the AI needed more prompting.
“It gave us a location that did not exist,” Eifert said. “So then, you know, you have to go through, and say, “Okay, write a tweet about Wright State’s library, and talk about the study areas or something like that.” Sometimes the team found it was easier to edit the AI’s work themselves which Eifert admits “honestly in a lot of circumstances took less time than trying to go through and say, “No, give me this.”
Knowles sees a future where AI levels the playing field for students who need help in improving their writing skills. Tom Webb sees additional benefits. He directs the Office of Disability Services at Wright State. Webb has cerebral palsy. He remembers in the 80’s when his teachers required him to use a typewriter. Then he learned about emerging speech to text programs.
Webb explained, “People had no idea or concept of what that technology could provide for students with disabilities, but I was using that very early on and saw how that was a game changer for me.”
Soon AI will take notes for students who are blind, deaf, or have physical disabilities. AI writing tools will also assist neurodivergent students who have difficulty structuring and ordering information. Knowles says that current AI writing tools are best at summarizing, outlining, and revising. Tom Webb imagines students will prompt the AI to provide sample exam questions based on class notes.“It allows them to know what to focus on,” Webb said, “and how to prepare for that exam. I think that right there is a huge tool, and I don’t see that being an unfair advantage.”
When Disability Services can automate many of these classroom tasks for their students, they can concentrate more on counseling and human relationships. Webb says that then “We get into more internships, clinical settings, building a student up to the point where they’re entering their career, what can we have in place for them to be successful.”