In some of McCormick sophomore Christa Kulanda’s STEM classes, she has noticed professors giving more weight to exams than before, citing students’ rising use of artificial intelligence. Kulanda said the change makes her feel like her “destiny was set in stone.” 

Many STEM classes assign problem sets, which are designed to encourage practice on concepts learned in class and typically account for a portion of students’ final grade. 

Kulanda said some of her professors have used exams for nearly the entirety of students’ final grades because of the increased use of AI chatbots like ChatGPT or Google Gemini to solve problem sets. 

“It just felt like the one opportunity that I did have to redeem myself or to really just feel like I could thrive was taken away from me,” Kulanda said.  

Economics Prof. Mark Witte said he has recently weighed in-class examinations more heavily in students’ grades. 

Witte added he has also encouraged students to work in groups on quizzes in Canvas to discourage AI use. But although he encourages collaborative work to avoid turning to AI, Witte said it’s not the perfect solution.

Compared to past years, he said more students answer online quiz questions right on the first try — and he doesn’t know if it’s AI-related or not.   

“It could be that people are being much more cautious and working out the problem more carefully before they try it the first time,” Witte said. “But it could also be that people are just plunking into an LLM and saying, ‘Hey, what’s the answer?’ and then putting that in.”  

Kulanda said that specifically in McCormick, she’s experienced some problem sets being worth fewer points as AI has increased in usage. 

Some professors are losing sight of the fact that students who can’t necessarily perform in high-stakes exam environments still use problem sets to show what they know, she said. 

“It’s pretty inaccessible and inadvertently exclusionary to people who may have conditions that inhibit them from really relying on the exams to make up their entire grade,” Kulanda said. 

To use large language models to benefit their education, Witte said, instead of asking AI for answers, students should ask these chatbots to critique their understanding of concepts or create practice questions.  

Due to students’ AI use, Weinberg first-year Siri Doddapaneni said quizzes that professors previously allowed students to take at home are now required to be taken in class. 

Although Doddapaneni said she appreciates how in-class quizzes prepare her for exams and reduce her cramming, the newly heavily weighted tests still don’t always accurately reflect her knowledge.   

“I feel like the stress and the environment of taking exams is not as conducive to my learning as having my own time to complete a problem set and truly take my time to understand the actual concepts,” Doddapaneni said.   

Still, some professors haven’t changed their grading styles. 

In her material science class, Kulanda said problem sets’ weight in grading haven’t changed. She said her professor provided the worked-out answers to questions, so students didn’t turn to AI first.  

Kulanda said her professor encourages students to show their work and learn from the answers he makes available. 

Economics Prof. Chiara Sotis said her grading style hasn’t changed much due to AI.

With her problem sets designed to incentivize effort, rather than correctness, Sotis said she hopes to give students the opportunity to try out problems on their own without fear of a strict penalty.

“The part that I care more about is convincing people that AI can be really valuable but shouldn’t be used at every stage,” Sotis said. “There’s a lot of learning that we give up if we’re just randomly inputting stuff and getting the output back.” 

Sotis said that many of her Microeconomics students have used AI to generate new practice questions based on past problem sets. 

Although she finds it to be a helpful tool, Sotis said that if students are looking to AI simply for solutions, the knowledge simply won’t stick.  

“I can see a thousand uses that seem completely fair and actually quite helpful,” Sotis said. “It’s more you would want it to help your learning rather than substitute your learning.” 

Email: [email protected] 

X: @alexingvold

Related Stories:

— Academic burnout is on the rise in higher education. What can students do?

— Feinberg Institute for AI in Medicine progresses computational medical research across campuses

— Northwestern experts discuss flexibility navigating AI in education