Petra De Sutter, who was elected as the new chancellor of Ghent University last year, used generative AI when drafting her opening speech in September, discovered journalists from Apache.
The software used by De Sutter “hallucinated” certain information, leading to incorrect quotes from Albert Einstein, among others. The university has since acknowledged and corrected the errors.
"Dogma is the enemy of progress" or "Knowledge describes and controls reality; wisdom seeks answers to existential questions – and is connected to culture and ethics" are just a few examples of quotes that appeared in De Sutter's speech. The first was attributed to German physicist Albert Einstein, the second to psychologist Paul Verhaeghe.
After the speech, however, several colleagues at Ghent University pointed out that De Sutter's text contained errors. The quotes turned out to be so-called "hallucinations," a known problem whereby information is fabricated by AI.
However, De Sutter herself did not spot the errors, and they ended up in the final version of the speech.
'Not blindly trusting AI'
Following these comments, Ghent University made initial corrections to the transcript of the speech published on its website. However, these corrections also contained errors. It was only after additional questions from the Apache editorial team that a disclaimer was added.
What's striking is that De Sutter herself referred to the dangers of AI in her speech. She warned that we should "not blindly trust" the output of AI tools and that AI-generated texts "are not always easy to distinguish from original works."
She also said that "it is the user of an AI tool who is responsible for the output, even when it generates incorrect or misleading information."
This is also evident, for example, in the examination regulations of Ghent University: students who are caught using AI when it is not permitted risk severe sanctions. These can range from a 0 on the course component in question to a ban on participating in resits and even, in serious cases, expulsion from the university.
In any case, the chancellor herself does not need to worry about this. The Ghent University Scientific Integrity Committee has not yet received any complaints.
'Fell into the trap'
Additionally, the rules governing the use of artificial intelligence in academic circles still seem to be in full development and vary depending on the context in which AI tools are used.
"An opening speech is not the same as an academic paper," Ghent University said in a response to Belga News Agency. This is also confirmed by Bert Seghers, an expert in scientific integrity. "When applying AI in scientific contributions, different expectations apply: being transparent about its use, taking responsibility for your work and correcting any errors. But a paper is not necessarily the same as a speech."
Still, Seghers cannot deny that De Sutter has an exemplary role to play. "The chancellor is the face of the university and, in that role, is also the guardian of academic integrity at the higher education institution. From that perspective, her actions are inevitably judged more critically and measured against the standards that Ghent University itself applies to its own students."
Ghent University also confirms that it expects both staff and students to "continue to critically examine the output" of generative AI.
In a response to Belga News Agency earlier on Thursday, De Sutter said she "regrets that she fell into the trap" and that she will use the experience to "further stimulate the debate on the use of AI."
