A recent MIT study shows artificial intelligence may be affecting our ability to think critically and problem-solve.
As AI continues to seep into classrooms and the workplace, it could come at a cost as faculty seek to find the dividing line between using AI to think with you, as opposed to letting it think for you.
Veljko Dubljevic, professor of philosophy and science, technology and society, said artificial intelligence can both be used to aid learning as well as hinder it.
“It’s usually the case of whether you know what you’re doing or you don’t know what you’re doing,” Dubljevic said. “How this is going to change cognition for people who already know what they’re doing is going to be amazing … for people who are just in training to learn how to do things, if they learn the incorrect way, it’s going to bite them in the back.”
Dubljevic added that there is a clear difference between using AI to help facilitate students’ learning and using AI to do one’s work for them, resulting in academic misconduct.
“If a student has read some sources and some of that is confusing and they have some bullet points they need more clarification on, then running that and asking the AI for clarification may be useful, right?” Dubljevic said. “But if you didn’t read it yourself and you just ask it to the AI and are like ‘hey, produce an essay,’ … it robs you of the opportunity to learn things.”
Noboru Matsuda, associate professor in the Department of Computer Science, stated that education is critical to ensuring proper use of large language models such as ChatGPT.
“We have to educate the students to know how to use AI for themselves, right? To actually help them learn LLMs,” Matsuda said.
A team of several professors, including Matsuda, is working on developing AI-powered tutoring tools to help students learn more effectively.
These tools can often be better tailored to students’ individual learning needs, Matsuda said.
“We have technology to compute the program probability that a particular student answers this particular question correctly, and then we compute the optimal question sequence,” Matsuda said. “Some researchers also apply AI techniques to know the next best kind of scaffolding, like ‘should we ask students another question, or should we show students an example?’”
Additionally, these tools are able to optimize content and instruction differently to separate students to better facilitate learning, Matsuda added.
Matsuda added there are concerns students may be using AI to replace their own thinking and to complete their work for them, instead of actually engaging with content.
“It’s not only saving them time but more like easing their pain. Instead of committing to challenge the tough questions by themselves, they ask AI to do so for them,” Matsuda said. “No pain, no gain.”
Matsuda said outright prohibition of AI across the board is not a good idea, as it can be a useful tool for learning if used properly.
Disclosure is also a critical element of AI usage. Dubljevic compared it to disclosing what programs are used to process statistics in research. He added that while many are hesitant to disclose when they use AI, transparency and academic honesty are critical in assignments and in research.
AI may have a place in certain classrooms but not others, Dubljevic emphasized, such as when developing fundamental skills.
“If the courses teach you certain basic skills, then you need to rely on yourself. Then there’s courses for more advanced students, when you already know certain things, then to use AI to improve what you’re doing, how you’re doing it, and so on,” Dubljevic said.
Ensuring AI follows human ethical frameworks is key, Dubljevic said. Dubljevic is the program director of the Centering AI in Society and Ethic Initiative.
Dubljevic also cited a 2025 research paper he authored, emphasizing applying the Agent-Deed Consequence Model to artificial intelligence systems to ensure ethical decision-making in AI systems such as autonomous vehicles, smart cities, or generative AI such as ChatGPT.
“AI is already making decisions that are affecting humans,” Dubljevic said. “So we need to equip them with a functional equivalent of human moral decision-making. So what we do is we explain how humans do it with the subcomponents of morality.”
The Neurocomputational Ethics Research Group, of which Dubljevic is principal investigator, will also be holding a conference in early June, aiming to discuss frameworks for AI ethics at NC State and beyond, with students invited to attend.
Dubljevic stated AI also holds promise in various fields, such as assistive medical care.
“Social robots are supposed to help people with mild cognitive impairment. So people whose cognition is starting to slip, AI can provide them with reminders, right?” Dubljevic said. “AI is providing you with the reminder and then you continue functioning relatively normally in your home, which is extremely important.”
Despite concerns around improper use, Matsuda remains optimistic about the future of AI in education and learning.
“I believe AI has a strong potential to make education better,” Matsuda said.
