The artificial intelligence (AI) file is essential this fall. AI has already changed, and will continue to change, many things in education. But only a smart person can predict what will happen, because nothing here is very clear.
Neither in education nor elsewhere.
Big questions
Faced with this big issue, I naturally tend to move towards more philosophical questions. For example: is AI artificial intelligence, or is it rather natural stupidity, as many people, including Chomsky, maintain? He notably argues that AI cannot think counterfactually, that is, in a hypothetical scenario. He is right, I think.
ChatGPT can simulate counterfactual reasoning, that is, imagining and describing hypothetical or alternative scenarios in response to given questions or situations. However, it is important to note that ChatGPT does not “think” in the human sense. It generates answers based on statistical models trained on a large corpus of texts, but it does not possess consciousness or real understanding of the concepts. In short, ChatGPT can indeed produce counterfactual answers, but these are text generation processes, not conscious thought.
Let’s move on. There’s also the whole terrifying question of what AI means for our future. Is it a serious threat to humanity, as Yuval Noah Harari argues, who sees it as a powerful tool of mass destruction? But perhaps it’s the “singularity” advocates, in particular, who are right when they say that it’s bringing us closer to the moment when humans will become immortal. The first immortals, they claim, are already among us right now.
But let’s get back to the real world, that of schools, colleges and universities. There, the questions raised by AI are linked to everything related to learning, research and teaching. Many and often surprising experiments are underway. I will come back to this. For the moment, let’s focus on cheating.
Cheating in the age of AI
The subject rightly concerns the Minister of Education, Bernard Drainville. From elementary school onwards, it seems, ChatGPT is used to cheat and do homework. It can only get worse over the years.
What to do? Right now, there is a lot of commotion about this. We are changing the way we assess, we are making students do their work in class, we are explaining to students what to do and what not to do with AI, and why. Not forgetting to equip ourselves to detect cheating, for example with the Compilatio software, which is said to be able to compare a text with millions of contents and detect cheating by identifying the source.
But here is a very recent article (June 2024) in which the authors report having submitted answers written by AI as part of entrance exams for a psychology degree at the University of Reading. The exam markers are not aware of this.
Overall, 94% of the AI-submitted responses were not detected, even though the AI had been used in the most detectable way possible. “We found that in 83.4% of cases, the scores obtained by the AI responses were higher than those obtained by a random selection of the same number of student responses.”
This dismal conclusion is added to others that also show that cheating detection software is not very reliable. Worse still, and this fact is also documented: not only are teachers bad at detecting cheating, they are also overconfident that they can do so.
What to do? I obviously have no magic solution and especially no criticism of the efforts that are being made in the face of a truly complex and difficult to grasp problem. But I am putting forward an idea, which research supports.
Here it is. Cheating can be reduced by reminding students what intellectual honesty means, how it translates into practice, and why it is important. Here is a interesting result published in the journal Contemporary Educational Psychology which demonstrates this. The authors tell us that “reminding students of academic integrity policies led to less cheating than no reminder; reminding students of actual instances of cheating led to less cheating; reminding students of the negative consequences of cheating if caught led to less cheating.”
In short: a code of honor stated and recalled seems to have its effect.
*****
A little gift
I offer you this one-page summary of the important theory of cognitive load. It is the work (in English) of Jamie Clark, an author and educational consultant who focuses on research-based teaching practices.
It was translated and laid out by Isabelle Lemieux, who teaches in 4e year at La Visitation Elementary School (CSSDM). She is preparing a master’s degree in teaching and school effectiveness at TELUQ.
Many thanks to both of you.
A naughty little question to finish
In the text you just read, an entire paragraph was generated by ChatGPT. Can you guess which one?
The answer next time.