With AI making inroads into classrooms and homework assignments, students now require learning how to reason more than ever before.
More students are relying on artificial intelligence for assistance with their assignments, causing educators to hurriedly modify their instructional methods or engage in discussions about prohibiting it entirely….
However, one professor compares AI to the introduction of calculators in classrooms, suggesting that the key is to teach students how to think critically about various problem-solving approaches and demonstrate when AI might mislead them.
In the last couple of years, Melkior Ornik, an assistant professor in the Department of Aerospace Engineering at the University of Illinois Urbana-Champaign, mentioned that both fellow educators and pupils have been concerned about numerous students utilizing AI models for their homework tasks.
Therefore, Ornik along with PhD student Gokul Puthumanaillam devised a strategy to determine whether such an idea could actually work.
Ornik clarified, “‘What we discussed was, let’s suppose that certain students aim solely for top grades or try to achieve an A without having any real understanding. Is this something they could realistically accomplish?'”
The scholars conducted a trial run within one of Ornik’s classes—a third-year undergraduate course focused on the mathematics of autonomous systems—to evaluate how well a generative AI model performed on coursework and tests.
The findings are detailed in a preprint document titled, “The Lazy Student’s Fantasy: ChatGPT Successfully Completing an Engineering Class Unassisted.”
Following our idea of mimicking the actions of an ‘extremely idle student’ aiming to pass the class without exerting much effort, we utilized the basic free edition of ChatGPT,” explained Ornik. “Generally, it did quite well, earning a modest B- grade for the course.
However, the AI model’s effectiveness changed depending on the kind of task assigned.
Ornik clarified, “One aspect that caught my interest regarding future adaptation is that although it performed reasonably well overall—scoring a B—it still showed considerable variation in handling various problem types.”
When dealing with closed-form problems such as multiple-choice questions or performing calculations, OpenAI’s ChatGPT, particularly GPT-4, performed very effectively. It achieved nearly perfect scores on these types of questions.
However, when more profound contemplation was needed, ChatGPT performed inadequately.
“Questions that were more along the lines of ‘go ahead, attempt to solve this issue and present possible solutions, followed by demonstrating with graphs whether your approach proves effective,’ performed much poorer,” stated Ornik. “In such tasks which we refer to as ‘design projects’, they received grades around the D mark.”
In Ornik’s view, the outcomes provide direction on how educators can modify their teaching methods to accommodate the anticipated utilization of AI in assignments.
He contends that the circumstances nowadays are similar to when calculators were introduced into educational settings.
Ornik pointed out, “In the days before calculators, individuals used to perform these trigonometric operations.” They relied on books containing logarithmic and trigonometric tables which stated things like, ‘if you want to find the sine of 1.65, go to page 600.’ However, as time passed, this method fell out of favor, and educators ceased instructing students in their usage since something more powerful had emerged: the calculator. Though perhaps not flawless, it proved sufficiently reliable. Consequently, they decided to place their faith in this new device.
So the genuine issue I wish to address—and one where I must admit I lack a definitive solution—is what subjects deserve our attention? Should we keep imparting knowledge in areas that AI can handle nowadays merely for the sake of enhancing students’ mental capabilities?
Or perhaps we ought to abandon certain aspects of this and concentrate instead on those higher-order questions that may not readily yield to AI solutions? And I doubt there’s yet a unified agreement on this matter.
Ornik mentioned that he has discussed with his colleagues at the University of Illinois’ College of Education the reasons behind teaching elementary school students mental arithmetic and requiring them to memorize multiplication tables.
Well, this is beneficial for their brain development even though they will eventually use phones and calculators,” he stated. “Teaching them these skills remains valuable for enhancing their future learning abilities and cognitive capacities.
I believe we need to engage in this discussion. What exactly are we instructing, and what is our rationale for teaching these subjects in an age where advanced AI is widely accessible?
Ornik mentioned that he envisions three approaches to tackle this problem. The first strategy involves treating AI as an opponent and structuring lessons so as to prevent the utilization of AI tools. This could entail implementing oral examinations along with tasks specifically crafted to be challenging to accomplish using artificial intelligence assistance.
Another approach is to view AI as an ally and merely instruct students on utilizing AI effectively.
“Then there’s the third option which is perhaps the option that I’m kind of closest to, which is AI as a fact,” said Ornik. “So it’s a thing that’s out there that the students will use outside of the bounds of oral exams or whatever. In real life, when they get into employment, they will use it. So what should we do in order to make that use responsible? Can we teach them to critically think about AI instead of either being afraid of it or just swallowing whatever it produces kind of without thinking.
The issue lies in how students overly rely on computational tools. We ought to emphasize using AI judiciously while ensuring that the information provided is accurate.
It may appear early to consider AI a certainty without having a business model that ensures its sustainability since AI companies currently operate at a loss, spending more than they earn.
Ornik admitted this was true, stating that he isn’t an economist and thus cannot forecast future developments. He likened the current situation to the dot-com bubble circa the year 2000.
He stated, “That’s definitely the sentiment we have nowadays with every product claiming to use AI.” He continued, “I was browsing barbecue grills recently—apparently, these barbecues come equipped with AI. It doesn’t quite make sense to me. Based on my observations, this seems to be using tech similar to what we’ve had around for about three decades; they’re simply relabeling it as AI.”
Ornik likewise highlighted unaddressed issues concerning AI models such as data privacy and copyright matters.
While those issues get sorted, Ornik and a handful of colleagues at the University of Illinois are planning to collect data from a larger number of engineering courses with the assumption that generative AI will be a reality for students.
“We are now planning a larger study covering multiple courses, but also an exploration of how to change the course materials with AI’s existence in mind: what are the things still worth learning?”
He stated one objective is to create a sort of critical thinking component that educators might integrate into their lessons. This segment would last about an hour or two and would inform students, “there’s this amazing tool known as ChatGPT. It possesses certain abilities, yet it can also fall short significantly. I’ll provide instances where it has fallen apart spectacularly, particularly those relevant to our current coursework.”
A further objective is to explore modifications in student evaluations and course content to accommodate the influence of generative artificial intelligence.
It’s quite probable that some courses might require being taught differently at times, and occasionally the content may remain valuable while we alter the assignments,” Ornik stated. “And then there could also be moments when one wonders whether certain topics deserve to continue being part of our curriculum.
