As the ethics of and practical uses of artificial intelligence (AI) are called into question in implementation in higher education, political science Professor Barbara Trish and anthropology Professor and Department Chair Monty Roper shed light on the moral dilemmas, strengths and weaknesses of AI’s position in the college classroom.
Trish, who specializes in American politics, is currently teaching a first-year tutorial class on the politics of AI. For the past few years, she has been studying the interference of AI in party politics, after completing a research fellowship. She is currently writing a book on the subject.
“Interference in campaigns goes on all the time with technology and without technology,” said Trish. However, when it comes to AI, she said she believes it takes a lot of resources to investigate who is and how they are trying to interfere into campaign politics.
Trish said the objective of her course is to make students think critically about AI. “It’s not unusual to have this great enthusiasm about something as possibly the fix for all sorts of problems,” said Trish. “Students must ferret out what’s really different and what’s a version of what’s already in place.”
Trish also emphasized the importance of thinking about the ethics of AI. One such concern is the human biases encoded into these models through the man-made information it is trained on.
“Visual models do much better with white individuals than with people of color,” said Trish. “Things that are really clear evidence of the fact that you should be concerned about bias. Then bring in the bigger questions about what if we get in a situation where the AI is making the decision or recommending decisions for us.”
She said she further expressed concern on the presence of misinformation and deepfakes and their effect on political discourse online. “The mere idea that there are fakes that are easy to produce, might make people question the truthfulness even of stuff that’s really true,” she said.
For students to succeed in a world in which AI is a commonality, Trish said she believes students must learn how to use it effectively.
“What I don’t know is whether, at Grinnell, we can directly prepare students to use AI in their work, or whether what we should do instead is continue teaching the strong general skills we’ve always emphasized—like critical thinking, writing and analysis—and then let students apply those skills to AI contexts on their own,” Trish said.
Meanwhile, Roper said he was inspired to test the capabilities of AI in academic research after attending a workshop.
“I was curious to know whether it was a viable research tool for me and to be able to understand what it can and can’t do in order to better understand how I could advise students into how it might be used effectively to do research,” said Roper.
Roper said this opportunity came during a summer research project, when he and a Mentored Advanced Project (MAP) student were analyzing the results of a qualitative study. Having to go through multiple transcripts traditionally using the qualitative analysis software NVivo, Roper decided to spend some of his research funds for Open AI’s most advanced generative language model, ChatGPT Pro 5, and test how it compares to the work of his assistant.
However, Roper said he ended up struggling to get the AI model to give him the results he wanted. Throughout his research, Roper said his interactions with AI sometimes involved arguing with it like a person. “I had a really kind of abusive relationship with AI over the summer where it abused me a lot, and I tried to abuse it back sometimes,” said Roper.
While the generative language model was effective in recognizing patterns and common themes in the transcripts, it struggled to pinpoint the evidence that led it towards its conclusion, according to Roper. It would also mismatch pieces of information and its source transcript or generate phrases that resembled the original quote, Roper said.
Eventually, Roper said that he and the MAP student came to virtually identical results, but he said that the time he spent correcting the AI model, drafting up more effective prompts, sometimes feeding back days of conversations, has cost him way more time than was necessary for the student. “It took me probably five to 10 times longer than if I just did it by myself,” he said.
Despite everything, Roper said his “research within a research” did not discourage him from believing in the future capabilities of AI. “AI is here and it is becoming big,” he said. “Things we thought we knew about it are not true six months later.” However, when it comes to students, he sees little benefit besides computing basic arithmetic and summarizing optional readings.
“I believe that students need to learn how to analyze data and think critically by themselves.” said Roper. “The comparative advantage of coming to a place like Grinnell is to learn to be a critical thinker, because that will survive across different jobs.”
