AI plagiarism hard to detect, harder to punish

Jan 18, 2024 | Campus News, News

Artificial Intelligence is a powerful tool and it is here on Humber Campus to stay.

However, Humber students and staff struggle to figure out how to accurately detect AI and academic dishonesty or even if it would be punished as a violation of academic integrity.

Humber’s academic integrity division said the college “embraces the integration of AI generative tools in ethical, equitable, and constructive ways in support of teaching and learning.”

All statements in the post, however, are vague and not policy.

Humber nursing Professor Jennie Miron, who has a doctorate in academic integrity, is also a board member of the International Centre for Academic Integrity and the chair of the Academic Integrity Council of Ontario, among other integrity committees. She is also the college’s Academic Integrity Lead.

“At Humber, there is no plagiarism detection software that we use connected to artificial intelligence,” she said. “The literature will tell you that it’s not effective and that you can actually have lots of false positives.”

Indeed, false positives are a significant issue in testing for AI, according to one report. False positives and a lower degree of accuracy in general when testing the writing of people with Autism Spectrum Disorder or people who speak English as a second language.

“Institutions cannot currently say with any degree of certainty that results from AI detection are accurate and non-biased,” according to the report.

This is a well-known issue within the academic integrity community, Miron said.

“Chat-GPT just exploded, so we’re still learning as we go, and most generative AI is based on the English language,” she said.

Miron also cites a lack of clear and definitive research into a problem as an issue while tackling it.

“There hasn’t been a ton of research done, but it does make sense that if you have somebody who writes where English is not their first language that they will be disadvantaged,” she said.

Mike Dover, a business professor at Humber College, wrote Dante’s Infinite Monkeys: Technology Meets the 7 Deadly Sins, a book about the use of technology for personal gain.

“Chat-GPT writes like a stuffy know-it-all who is middle-aged but also comes across as a lazy but confident C- student,” he said.

Dover said a person’s writing style is easily identifiable and linguistically can even reveal their first language.

“Someone who speaks English as a first language will say three over seven, but someone who’s from India will say three on to seven,” he said as an example.

Dover states laziness is the leading cause of academic dishonesty, but Dover uses a more specific name.

“Academic dishonesty is mostly sloth out of any of the sins,” he said.

Dover said it’s hard to test for and to find people guilty of especially if teachers cannot be sure they cheated intentionally in the first place.

“You should not punish someone who does not have a guilty mind,” he said.