The development of ChatGPT and other artificial intelligence (AI) software programs means coursework and other forms of teacher assessment are no longer viable due to the risk of cheating by students, according to a new think tank report.
The report, Examining exams: Are there credible alternatives to written examinations? by EDSK (Education and Skills) suggested that while, in the past, coursework always carried a risk of malpractice, the rise of AI tools will make it almost impossible to spot or stop students from cheating.
“When you’ve got such powerful tools at the disposal of students in high stakes settings, such as A levels and GCSEs, then the risk of cheating just becomes insurmountable,” Tom Richmond, director of EDSK and co-author of the report, told Tes.
He added that because ChatGPT has the power to write entire essays or projects from a simple prompt, such as “write an essay on the role of nature in Frankenstein”, it will be almost impossible for teachers to spot plagiarism, and therefore undermine any grades given to coursework.
“We need to make sure that the grades are fair across the whole country, let alone within an individual school, that’s where we think ChatGPT really does present a very serious challenge that goes beyond what any individual school can tackle,” he said.
Although English is the most obvious area of concern, Mr Richmond noted that AI that can produce photographs and illustrations could even undermine art coursework: “I think a lot of people don’t realise the scope of power these types of programs possess.”
These concerns echo comments by Ofqual’s chief regulator, Dr Jo Saxton, who recently stated if she was a school leader she would make students do coursework under exam conditions to reduce the likelihood of cheating (reviving talk of controlled assessments), and that ChatGPT “reinforces the importance” of exams that “have stood the test of time so well”.
This view is at odds with the recent announcement by Matt Glanville, head of assessment principles and practice at the International Baccalaureate (IB), that students can use work generated by ChatGPT in essays as long as it was properly sourced.
“As with any quote or material adapted from another source, it must be credited in the body of the text and appropriately referenced in the bibliography,” he said in February.
“To submit AI-generated work as their own is an act of academic misconduct and would have consequences. But that is not the same as banning its use.”
Responding to the report, Dr Mick Walker, president of the Chartered Institute of Educational Assessors (CIEA), told Tes that while he agreed malpractice from AI was a concern, technology may evolve to get smarter at spotting it, too, which could help teachers.
“Although there isn’t a package yet to spot artificial intelligence, packages are being developed…I’m sure the technology on that side will run parallel with this.”
He also said it was important not to downplay a teacher’s knowledge of their own students as a means to combat the risk of malpractice.
“I could spot within 30 seconds when a child used to come into school and hand in a piece of homework or coursework that clearly had an adult input or somebody else’s,” Dr Walker said.
“If you know the child day-to-day in the classroom, you see what they write, you can spot it within two minutes.”
Finally, he noted that AI software is far from perfect and could easily lead to errors that teachers would quickly spot. “It often gets dates and facts wrong [but] the problem is most kids wouldn’t see that. But then it relies on the teacher spotting those deficiencies,” he said.
Rethinking exams
Nevertheless, given the concerns with AI, the EDSK report suggested that ”written examinations should continue to be the main method of assessing students’ knowledge and understanding” as the most straightforward way of ensuring the credibility of the work students are producing.
However, despite this, the report did acknowledge that written exams themselves are not perfect, and can struggle to ensure that students develop a broader range of skills that can help them in higher education or employment.
One example of where the report claims exams fall short is the development of students’ verbal communication skills, which evidence suggests employers frequently complain that school, college and university leavers are often lacking.
“We do think there are some important reforms that need to be made to introduce a more varied mix of assessments, as long as those assessments can withstand the pressure of a high-stakes system,” Mr Richmond added.
Through its analysis of other forms of assessment, the report found that oral exams - long used for assessing subjects such as language learning - could be a suitable alternative to written exams.
“A student going through a two-year A level and doing nothing but written exams in a silent exam hall on pen and paper will have significant downsides,” Mr Richmond said.
“Written and oral exams are both able to operate in a high-stakes assessment system. They both develop very different skills among the students.”
