By Barbara Lewis and Supantha Mukherjee
LONDON/STOCKHOLM (Reuters) – At leading Swedish university Lund, teachers decide which students can use artificial intelligence to help them with assignments.
Advertisement
At the University of Western Australia in Perth, staff have talked to students about the challenges and possible benefits of using generative AI in their work, while the University of Hong Kong is allowing ChatGPT within strict limits.
Launched by Microsoft-backed OpenAI on Nov. 30, ChatGPT has become the world’s fastest growing app to date and prompted the release of rivals like Google’s Bard.
GenAI tools, such as ChatGPT, draw on patterns in language and data to generate anything from essays to videos to mathematical calculations that superficially resemble human work, spurring talk of unprecedented transformation in many fields including academia.
Academics are among those who could face an existential threat if AI is able to replicate – at much faster speeds – research currently done by humans. Many also see the benefits of GenAI’s ability to process information and data, which can provide a basis for deeper critical analysis by humans.
“It can help the students to adapt the course material to their individual needs, aiding them much like a personal tutor would do,” said Leif Kari, vice president for education at Stockholm-based KTH Royal Institute of Technology.
The United Nations Educational, Scientific and Cultural Organization (UNESCO) on Thursday launched what it is says is the first global guidance on GenAI in education and academic research.
For national regulators, it outlines steps to take on areas such as data protection and revision of copyright laws, and urges countries to make sure teachers get the AI skills they need.
CHEATING VERSUS HELPFUL SHORT-CUTS
Some educators draw a comparison between AI and the advent of hand-held calculators, which began entering classrooms in the 1970s and stirred debate on how they would affect learning before they were quickly accepted as essential help.
Some have expressed concern that students might similarly rely on AI to produce work and effectively cheat – especially as AI content gets better with time. Passing off GenAI as original work could also raise copyright issues, prompting questions over whether AI should be banned in academia.
Rachel Forsyth, a project manager in the Strategic Development Office at Lund University in southern Sweden, said a ban “feels like something that we can’t enforce”.
“We’re trying to put the focus back on learning and away from cheating and policing the students,” she said.
Worldwide, the software Turnitin has for decades been one of the main ways to check for plagiarism.
In April it launched a tool that uses AI to detect AI-generated content. It has provided that tool free to more than 10,000 education institutions globally, although it plans to charge a fee from January.
So far, the AI detection tool has found that only 3% of students used AI for more than 80% of their submissions and that 78% did not use AI at all, Turnitin data shows.
Problems have arisen over what are known as false positives when text written by humans – in some cases by professors trying to test the software – has been flagged as written by AI, though those wrongly accused of using AI can defend themselves if they have saved various drafts of their work.
Students themselves are busy experimenting with AI and some give it a poor grade, saying it can summarise at a basic level, but that facts must always be checked because GenAI cannot distinguish fact from fiction or right from wrong.
Its knowledge is also limited to what it can scrape from the internet, which is not enough for very specific questions.
“I reckon AI has a far way to go before it’s properly useful,” said Sophie Constant, a 19 year-old law student at England’s University of Oxford.
“I can’t ask it about a single case. It just doesn’t know and it doesn’t have access to articles I am studying so it is not very helpful.”
CORPORATE SPEED AND SLOW-MOVING REGULATION
UNESCO’s latest guidance also flags the risk GenAI will deepen societal divisions as educational and economic success increasingly depend on access to electricity, computers and internet that the poorest do not have.
“We are struggling to align the speed of transformation of the education system to the speed of the change in technological progress,” Stefania Giannini, assistant director-general for education at UNESCO, told Reuters.
So far, the European Union (EU) is among those at the forefront of regulations around the use of AI with draft legislation that has yet to adopted as law. The regulations do not specifically deal with education but its broader rules on ethics, for instance, could be applied to the field.
After its exit from the EU, Britain is also trying to work on guidelines for the use of AI in education by consulting educators and says it will release the results later this year.
Singapore, a leader in efforts to train teachers on how to use the technology, is among the nearly 70 countries that have developed or planned strategies on AI.
“In terms of universities, as a professor, rather than fighting it, you need to leverage AI, experience it, develop a good framework, guidelines and a responsible AI system, and then work with students to find a mechanism that works for you,” said Kirsten Rulf, a partner at Boston Consulting Group.
Rulf co-negotiated the European Union AI Act in her previous role as head of digital policy at the German Federal Chancellery.
“I think we are the last generation that has lived in a world without GenAI.”
To hear about how educators are tackling the advance of AI in the classroom, click here for the Reuters World News daily podcast.
(This story has been corrected to clarify that Rulf’s previous role was at the German Chancellery in paragraph 27)
(Editing by Deepa Babington and David Goodman)