Clones in the classroom: why universities must be wary of embracing AI-driven teaching tools

The university sector in Aotearoa New Zealand is at a tipping point due to chronic underfunding, shifting enrolments and increasing costs from inflation. In response, the government has established two working groups to assess the health of the sector and provide recommendations for the future.

Meanwhile, universities find themselves increasingly beholden to the education technology (EdTech) industry, which claims to improve student learning by selling hardware and software – often built with artificial intelligence (AI).

Most universities already pay for services from EdTech companies such as Turnitin, Grammarly, CampusTalk and Studiosity, all of which use AI in their products.

But critics say this trust in EdTech is misplaced and amounts to what technology writer Evgeny Morozoc calls technological solutionism – “the idea that given the right code, algorithms and robots, technology can solve all of mankind’s problems”.

Intellectual property and profit

To better understand how EdTech providers work, consider the plagiarism-detection company Turnitin, used by 20,000 institutions in 185 countries.

Student essays are intellectual property (IP). According to university policy, students and universities have a shared licence over any IP written by students.

But when teachers require students to upload their essays to Turnitin, students must grant the company a “non-exclusive, royalty-free, perpetual, worldwide, irrevocable licence” to use their IP.

Turnitin adds those essays to its massive database, which it uses to build its plagiarism detection tool. The tool – built with student IP – is then sold back to universities so educators can police student writing in the name of academic integrity.

After 20 years of growth with this business model, Turnitin was bought in 2019 by Advance Publications for US$1.75 billion. On Twitter (now X), higher education researcher Jesse Stommel asked: “How much of that $1.75B do you think is going to the students who have fed their database for years? I have a pretty good guess: zero billion.”

By claiming ownership over student IP, Turnitin also profits from Indigenous students’ ideas. But this threatens Indigenous data sovereignty – that data produced by or about Indigenous communities should be governed by those communities.

AI teaching clones

EdTech organisations such as Prifina, Khanmigo and Cogniti are now developing new AI teaching clones. These “AI twins” are trained on educators’ own course materials and can interact with students around the clock.

For overworked teachers, an AI clone might seem appealing. In one promotional video, a lecturer praises the clone for helping him teach biochemistry to more than 800 students.

Of course, another way to improve the teaching of such a large course is to hire more teachers. The Tertiary Education Union will surely emphasise this point in its collective bargaining with universities this year.

But it’s not surprising universities are looking for cheaper options, given the sector has endured long-term government underfunding.

Here’s the catch, though: we don’t yet know the full cost of AI agents in education. They may be free or cheap now, but it takes a lot of computing power to create and engage with a customised AI agent – almost certainly more power than teachers and students have on their personal computers.

For that reason, organisations that develop AI agents rely on access to high-performance servers provided by the likes of Microsoft and Amazon Web Services. After sufficient market penetration, though, these multinational companies will want a return on their investment. Will AI agents still be cheaper than teachers then?

Energy-intensive investment

We also don’t know how much energy it takes to build and operate hundreds of AI clones for tens of thousands of students. But we do know Microsoft recently retracted its goal of being carbon-negative by 2030 due to AI’s increasing demand for “energy-hungry data centres”.

Many New Zealand universities also aim to be carbon-neutral by 2030. Will they too have to renege on their green commitments? Or will outsourcing AI to EdTech companies allow them offload responsibility?

Historically, it has been difficult for teaching institutions to untangle themselves from EdTech investments. This is despite research showing “85% of EdTech tools are poor fits or poorly implemented”.

If AI is “pushing the world toward an energy crisis” is it worth the financial and environmental cost to create AI agents for educators?

Product or public good?

Without sufficient government funding, EdTech products look appealing to universities. But tertiary stakeholders must question whether EdTech “solutions” really contribute to a university education.

If a university education is primarily viewed as a product in a global marketplace, then EdTech tools might add monetary value.

But if a university education is viewed as a public good that contributes to the improvement of society, then EdTech tools might be less valuable.

Now is the time for a broader conversation about the cost and value of a university education, and the role of EdTech within it.