What will ChatGPT and the next generation of AI tools mean for international education?

A new AI-powered chatbot, ChatGPT (from Silicon Valley start up OpenAI), is provoking heated debate within academic circles about how to handle the chatbot’s influence in the classroom and on exams.

ChatGPT responds to users’ queries by quickly searching massive databases – including datasets of real human conversations – to answer questions in a way that seems both eerily human and not at all human due to the speed at which the chatbot delivers results and answers.

An article in NPR sums up the reason for the commotion: “ChatGPT [has] superhuman abilities to solve math problems, churn out college essays, and write research papers.” In The Atlantic, English teacher Daniel Herman writes:

“The arrival of OpenAI’s ChatGPT, a programme that generates sophisticated text in response to any prompt you can imagine, may signal the end of writing assignments altogether—and maybe even the end of writing as a gatekeeper, a metric for intelligence, a teachable skill.”

Mr Herman says he has seen ChatGPT “draft a reasonable college essay, a cover letter to serve as a manager at Starbucks, and even an academic paper comparing two texts.”

New York Times writer Kevin Roose commented, “ChatGPT is, quite simply, the best artificial intelligence chatbot ever released to the general public.”

It may be incredible, but naturally people wonder if it is too incredible for humanity’s good. If it thinks for people, writes for people, and convinces people, what is left for humans to do but … plagiarise? Why would a student go to the extra effort of writing original thoughts if that student is time-pressed or otherwise overwhelmed?

The answer: many students will use ChatGPT to plagiarise, and many already are. Writing for Inside Higher Ed, Jeremy Weissman reports:

“A lecturer at an Australian university found that a fifth of her students had already used ChatGPT on their exams. Scores of Stanford University students reportedly used it on their fall 2022 final exams mere weeks after its release.”

What is to be done? Mr Weissman suggests a range of interventions that institutions should consider to cope with the issue – as coping is likely a more realistic strategy than any attempt to solve the problems unleashed by this accomplished chatbot.

These include designating an official AI task force at the institutional or even departmental level, banning technologies in classrooms (with some exceptions), relying more on in-class assignments, using Internet-disabled computers in classrooms, and more.

ChatGPT detectors are already in development, and these will be helpful to instructors faced with the problem of how to adapt their teaching and marking given the monumental disruption caused by ever-more sophisticated AI. But as Mr Weissman notes in his article (which relies on a COVID-19 metaphor for its impact):

“The efficacy rate of these GPT detectors may soon drop as a stronger mutation of the GPT virus emerges just around the corner: GPT-4—the Delta of GPT. And so the GPT detectors will be updated to protect against GPT-4, and the alphabet of GPT variants coming, but the efficacy may keep weakening, and the mutations will keep occurring. Unlike with COVID, the new variants of GPT will only keep getting stronger, even exponentially so.”

A powerful new tool for recruiters?

Sophisticated chatbots have already changed the way that schools and universities communicate with students, and they have been shown to boost conversion rates from 3% (without AI) to 8% (with it). In other words, if 3 of 100 students visiting an institutional website normally go on to enrol, AI can change that proportion to 7 or 8 who enrol as a result of visiting the website.

Some schools are using AI-supported admissions processes that crunch vast amounts of data to help admissions staff identify and focus on the candidates that are best fitted to the institution or school in question. And “best fit” in this sense can include sweeping considerations around not only which students are most likely to be successfully admitted, but also those that are most likely to secure a study visa, graduate, and even have a successful career after their studies.

Data show that this further application of AI can result in a higher graduation rate for AI-supported institutions (from 68% for “non-AI” institutions to 78% for those using AI tools).

It's reasonable to imagine that the emergence of more sophisticated AI tools such as ChatGPT may now open the door to improved machine learning-assisted student services, including greater responsiveness and effectiveness in responding to student queries.

Paradoxically, the sophistication and prevalence of AI may also lead to a greater need for technology-free assessment tools (e.g., in-person interviews) to discern whether students’ true capabilities match up with their grades and writing submissions.

For additional background, please see:

• “Data shows that machine learning can boost international admissions conversions”