The Starset SocietyUniversities are increasingly introducing artificial intelligence (AI) teaching tools into classrooms.
The latest case to make headlines is Macquarie University’s “Virtual Peer”—an AI chatbot introduced in two psychology units. In the online delivery of these units, some material previously taught through Zoom tutorials is now delivered by the AI chatbot and other online activities.
This comes as hundreds of University of Sydney staff walked off the job last week, calling for greater job security as AI teaching tools become more common. Meanwhile, some students have also raised concerns that replacing teaching interactions with AI could devalue their degrees. So is AI use in courses a good thing? And when might it become a problem?
How AI is being used in teaching
Universities are using AI for a range of purposes. From a student’s perspective, most uses fall into four familiar functions: tutoring, lecturing, feedback provision and marking. These are all core parts of university teaching traditionally carried out by human teachers. They are not new functions or abilities created by the availability of AI.
Among the most common uses of AI at universities is feedback provision. Educators at the University of Sydney, for example, have used the AI platform Cogniti to allow students to get 24/7 feedback on their assessments before submission.
Similarly, this year the University of Wollongong began trialing AI “study buddies” across 15 large first-year subjects. Their aim is to provide academic support to new students and answer their common questions as they transition into university.
Some systems extend tutoring into practice and rehearsal. La Trobe University’s “Fletch” plays a disgruntled parent, changing its responses as trainee teachers work through a difficult conversation.
Lecturing by an AI rather than by a human is less common, at least at the moment, but it is starting to appear. For example, the University of Newcastle has announced that an AI avatar will deliver lectures in a computing subject. This involves the academic writing the script while software generates the presenter’s voice and moving face.
Is this a bad thing?
From an institutional perspective, there are good reasons to develop these AI tools. First, well-designed educational AI might allow universities to better monitor and control students’ AI use. Research has shown that around half of Australian university students use AI on their own initiative for feedback on their work. This AI use, unless strictly controlled by the students themselves, is at least as likely to provide answers—or simply do the work itself—as it is to provide real feedback.
When universities create their own in-house feedback AI agents, they can aim to set better controls and boundaries around students’ AI use. This helps universities ensure AI use is supporting, rather than replacing, learning. Or, at least it can, as long as students limit themselves to using the AI their institutions provide.
Second, students may well benefit from the around-the-clock feedback AI can offer. Research has consistently shown that one of the biggest complaints students have about feedback is its lack of timeliness. When AI is used as a substitute for human-provided feedback, students can ask questions whenever they arise. They can also request additional explanations without embarrassment.
Third, AI tools can communicate with students in their own language. My own research suggests around half of Australian students use AI to read academic texts at university. The groups most likely to do so are students translating ideas and concepts into their native language or into a disciplinary context they are most familiar with.
There are risks
Of course, there are disadvantages to using AI tools in universities as well.
AI can be inaccurate, make things up entirely, flatten out complicated ideas and present concepts from a narrow Western male viewpoint. There are also the wider environmental concerns that come with AI use due to the energy consumption of data centers. The lack of rules around how universities should use AI in courses complicates how we should weigh these risks. The Australian government has regulations around teaching quality and student support. But there are no specific rules telling a university when, where, why or how an AI chatbot may or may not be used.
These are serious questions to resolve. Research shows the decisions we make now might be hard, even impossible, to undo. All these concerns matter. But even if they were solved, a different problem might remain.
What is lost if AI replaces a human educator?
Students need to be recognized
Education is more than just a simple transfer of information.
When a student asks a question in a tutorial, the answer helps them, but their question also shows the educator where the class is and what might need more focus. This interaction helps the educator tailor their teaching appropriately.
When feedback comes from an educator, it can also show the student their work and development are worth taking seriously. This is what my research terms “recognition.”
One of the most important things we risk losing when we replace a human educator with AI is our mutual recognition of each other.
Recognition is not simply praise, kindness or “human connection.” An educator might recognize a student by challenging a weak argument, noticing genuine progress or expecting more from them in their work. Such responses position the student as someone capable of participating in a field of knowledge and whose ideas deserve serious judgment.
An AI system can produce clear and useful feedback, and it might even generate the same words as an educator. But it cannot, by itself, recognize the student. The information may be similar, but that doesn’t mean the educational interaction is, too.
read more at phys.org
