During the holiday break, Jonathan Laberge, who teaches high school English, started to peg on ChatGPT. “I was reading stuff, it was just negative. I thought, “Surely there’s something to do with that!” So I asked myself, “What could ChatGPT do for me?” »
Jonathan Laberge first asked the OpenAI chatbot to write him a lesson plan based on themes he provided. Within seconds, it was all there. Then, he asked the robot to offer him discussion questions about the series Alice in Borderland, of which the students had watched a first episode.
ChatGPT was also able to easily produce quiz questions, which were then projected onto a classroom screen and answered by students using their computers. The questions were very well designed. Except two, which contained inaccuracies.
“Some questions were irrelevant. It clearly takes a teacher’s head to judge whether it’s right or not. The robot sometimes makes mistakes, and you have to know the material to detect them. »
Designing a post-apocalyptic type game where students have to discuss among themselves – in English, of course – the best items to buy to survive? ChatGPT produced the thing in minutes. “If I had done this myself, it would have taken me hours!” There, I had 30 minutes of really fun activity that I could never afford to do normally. »
Jonathan Laberge could have stopped his exchanges with the conversational robot there. The experience, for him, would have already been conclusive. He chose to push the machine even further. He asked his students to produce a critical summary of a work, and ChatGPT to provide him with a correction grid.
The grid provided by the robot was basic. Unsatisfactory.
It is at this point that we enter another world: the English teacher taught ChatGPT how to produce a correction grid to his liking. He first copied and pasted a criterion-referenced correction grid from the Ministry of Education as a model.
“I explained to him precisely what I wanted. And he did! It was all there! To be certain, I submitted this to the educational advisor of the CSSDM [centre de services scolaire de Montréal]. She said to me, “Come on, let’s see!” »
And then, the ultimate test, was the machine able to correct students’ critical summaries? Jonathan Laberge entered a student’s text into ChatGPT. At the teacher’s request, the machine underlined each of his errors, classified the errors by theme, and indicated to the student the grammatical rule that applied. “That’s sick!” »
ChatGPT’s final score for the student’s critical summary: 75%. Teacher’s final grade: 82%. “Let’s just say we weren’t a stratosphere apart…” The robot could even provide the student with tips on how to improve the quality of their work. Vary the sentence structures, give more concrete examples in the section on the strengths and weaknesses of the work…
In the space of two months, Jonathan Laberge went from a teacher who wondered what ChatGPT could do for him to a teacher who planned, “in a snap”, the activities of the whole month of February in his English classes in collaboration with the OpenAI robot.
“The answer to my original question is that there’s not much he can’t do!” »
What the teacher gains is time, testifies the teacher. “The time we can save is phenomenal. He’s a personal assistant who doesn’t growl behind your back at the coffee machine! he laughs.
And while he is not busy correcting, planning, imagining how to stimulate his students, the teacher can devote himself… to teaching.
Annie Dumay, pedagogical advisor in technology at the CSSDM, believes that artificial intelligence will become an assistant for teachers. The software will allow them to produce specialized and original content that goes far beyond textbooks.
Philippe, a colleague of Ms. Dumay, experienced this very personally after the holidays. On the suggestion of the educational advisor, he used ChatGPT to create tailor-made readings for his son Maxence. The little one is 10 years old, but, suffering from dyslexia, he is just beginning to read. He must practice… but the contents of the texts for learning to read, which are too “baby”, do not interest him.
“His teacher said he had a motivation problem. We wondered what we could do for him,” he says. The father had the idea of using the Playmobil figurines which are part of a daily family ritual. “He is extremely creative with his Playmobil men. »
Like Jonathan Laberge, Philippe started to peg on ChatGPT. He printed a first story. “And that’s when my son said, ‘Couldn’t the policeman in the story do such a thing?’ And, of course, it was possible. ChatGPT received new settings. The father printed his work, with a typeface and spacing suitable for dyslexics. He even asked the robot to produce images to illustrate the story.
“Within 10 minutes, I had a reading teaching tool for my boy. Not only did the theme interest him, but he contributed to the adventures of the characters. “The first weekend, Maxence read more than in his entire life, testifies the father. “As a teacher, I said to myself that there was something extremely interesting for students in difficulty, as well as for students in francization, for example. Philippe asked us not to reveal his last name so that his son would not be identified.
At the Laboratory for Innovation and Educational Research at Collège Sainte-Anne in Lachine, we are already working on the class of the future assisted by artificial intelligence. The three employees who work there are currently designing software that will allow French teachers to identify and classify their students’ errors, produce a diagnosis and suggest ways for each young person to improve their writing.
“And the evidence shows us that the faster the feedback time, the greater the improvements. However, teachers, and it’s not their fault, can take weeks to correct copies, “explains Yannick Dupont, who heads the laboratory.
“It would be an incredible time saver,” said the director of Sainte-Anne College, Ugo Cavenaghi. It’s not done: we are at the research stage. But if the teacher spends less time correcting, he will have more time to teach. We believe it: we took a French teacher out of a class this year to work on it. »
Already, the software has been fed with a dozen written productions by students. “We can know that behind the B obtained by such a student, it is on such and such a point that he makes his mistakes, underlines Mr. Dupont. It is very difficult to follow 120 students so closely for 32 criteria over a school cycle. »
“In fact, he adds, it is humanly impossible. »
“In education, the arrival of AI will be as destabilizing as the arrival of the internet in homes. »
Dave Anctil knows what he’s talking about. This philosophy professor at Jean-de-Brébeuf College is a researcher affiliated with the International Observatory on the Impacts of Artificial Intelligence and Digital Technology. For him, it is obvious that AI will completely change the game in education.
“An amazing number of new tools will be available, which will make it possible to make matter alive, to study in another way, he explains. It’s downright revolutionary, and I’ve seen it coming for a long time. »
An example ? A student could use a chatbot to “chat with Socrates” to revise material for an exam. “It will reassure the students: they will be able to ask their questions, which they might not dare to ask in class because they are afraid of looking ridiculous. It will stimulate them. »
Young people will also be able to access concepts and material much more advanced than their school level thanks to the popularization capacity of AI.
Because yes, conversational robots sometimes declaim bullshit. Students should therefore remain alert before using this information. “We have to train students in techno at school, we have to develop computational thinking, critical thinking about it. We will have to learn to be critical, to cross-check, “said Patrick Giroux, professor of educational technologies at the University of Quebec at Chicoutimi (UQAC).
Bots also have biases: ChatGPT, for example, is rather Eurocentric, Giroux points out. “From an African point of view, for example, there is a whole part of the story that is missing. Students will need to learn to identify these biases and work around them.
We will also have to train young people to interact with machines: more than ever, the art of the question asked will be crucial to obtain a good answer. “It’s all command-line art.” »
Just recently, Open AI released ChatGPT version 4, with even greater capabilities. Other digital giants have also introduced their own chatbot: Bard for Google and Bing for Microsoft.
The change in education will be profound, also predicts Mr. Giroux. “The school should have realized that declarative knowledge – in which year the Constitution was signed, in which year Jacques Cartier discovered Canada – should not be evaluated since the appearance of Google. If it’s googleable, it shouldn’t be in the review. This information is in everyone’s pocket and can be found in 30 seconds. »
Tasks performed in the classroom and at home are also likely to undergo a transformation, believes Dave Anctil. “Basically, it is to be read at home, and executed in class. This is the flipped classroom model: students absorb theory at home, and practice and assessments take place in the classroom.
Yes, it will be destabilizing for teachers, recognizes Mr. Anctil. But they will quickly realize how much time AI can save them by supporting them in tasks other than actual teaching. For him as for the other experts, it is obvious that teachers continue to play a crucial role in the classrooms.
“Professors must have the latitude to choose when and how technology is relevant. The question to ask is, is this going to hurt skills rather than build them? If it hurts, AI is counter-productive,” says Simon Collin, holder of the Interuniversity Research Chair in Education and the Teaching Profession, and professor at the University of Quebec in Montreal (UQAM).
And that’s why an institution like the Montreal School Services Center (CSSDM) has chosen not to ban the use of ChatGPT. “Rather than outlawing the use of chatbots like ChatGPT, we need to think about using them as learning tools,” says the notice sent to all directorates in February.
“We can’t ban AI. It would be like banning the internet: it is impossible. It’s a losing battle. The computer, we use it as a tool, we use the internet in an intelligent way: it is the dictionary or the Bescherelle of the modern world. It’s not tech for tech, it’s how we use it,” said Ugo Cavenaghi, director of Collège Sainte-Anne, a school where students have had laptops for 15 years.
In the public network, teachers have been pushed to update themselves in technological matters with the pandemic, observes Ms. Dumay. “We moved forward 10 years during the pandemic. Teachers had no choice but to learn to use digital tools. And after returning to class, they continued to use many of these tools. »
Are teachers ready for a new revolution?
Either way, it looks like it’s coming.
ChatGPT is just the tip of the iceberg. Artificial intelligence applications that could be used by teachers and students are multiplying. Here are some examples, already available or soon accessible, suggestions from teacher Dave Anctil, as well as from a Quebec site dedicated to digital education, Thot Cursus.
An AI for learning to read. Students read a text aloud and transfer the recording to the teacher. The app can analyze student progress. Accessible in French by Microsoft Teams, free of charge.
Personalized tutors in AI mode. From manuals, ProfessorBob.ai can generate thousands of exercises and activities. He can teach certain concepts to the students, have them revise them, answer their questions. French version offered. As for Ecree, “he is one of countless mentoring agents created for English-speaking academia and aimed at people who don’t have the money (and therefore less time) to further their education, thus contributing to the equal opportunity,” says Dave Anctil. The app can even help students better plan their schoolwork by interacting with their calendar.
A language model à la ChatGPT. “His particularity is not to do the work in place of the editor, but to increase his writing skills in English, by accompanying him, by supporting him continuously in his work”, underlines Dave Anctil, for whom this app is a great example of a successful approach to human-computer interaction. In English.
An application that optimizes the work of writing in English, “very useful for people who have to write in English without being English speakers”, says Mr. Anctil.
This portal provides access to a language model trained specifically in French. “He can be asked to do many things: write, revise, suggest corrections, analyze texts written by others… His fine knowledge of French from France makes him a versatile and precise tool,” says Mr. Anctil.
These two applications generate images from the commands given by the user, whether they are words or artistic or photographic styles.
These two apps can compose music according to a specific style, sometimes with given lyrics.
The app will answer questions posed by the user by drawing quotations from tens of thousands of books, whose references will be cited.
The first app reads documents for the user and summarizes everything for them in seconds. The second transcribes and analyzes audio and video content and summarizes them.
These apps create presentations, PowerPoint style, or tell stories on a given subject by automatically generating the text as well as the images. Wooclap is more focused on interaction with the audience.
Creates Excel or Sheets documents according to the instructions given by the user.
Two anti-plagiarism tools, which allow you to check whether a text has been composed… by an artificial intelligence. Obviously, the answer only comes in the form of probability.
Éric Martin and Sébastien Mussi, both CEGEP philosophy professors, are worried. And it is palpable in the book they have just published, Welcome to the machine. For them, screens have invaded the education sector in recent years, and certainly not for the better. Are we now headed for disaster with the arrival of AI in the classroom?
Sébastien Mussi: There was already a swelling that was present before COVID, but increased after COVID on the distance learning requirements. And we, what we have seen is that distance learning does not work well. However, we continue anyway. Computer resources are becoming requirements and screens are invading classrooms. In addition, all this is done outside democratic circuits. This book was born out of frustration with that.
Éric Martin: During the pandemic, we spent months in front of black screens in front of students who did not turn on their cameras, it was disastrous. And yet, government directives are to increase blended and distance learning! Studies show that it does not work, that it causes mental health problems, and yet we continue to press the accelerator.
EM: Yes. A good example of this is UQAM. Before the pandemic, there were about fifty online courses, it was very niche. Now there are 850! It didn’t go away with the end of the pandemic, on the contrary! There is a desire, both at the OECD and at GAFAM, to move forward with a tele-everything society, where everything is done remotely, including education.
SM: The only place that seems a little protected is the primary and secondary, because we saw how catastrophic it was. But even in primary and secondary school, computer work is increasing.
EM: I don’t think there’s any point in leaving training in the hands of GAFAM, so that the school and society in general are increasingly entrusted to algorithmic governance. This does not mean that AI is bad in all matters, but because it is developed in a capitalist society, it is very likely to participate in this phenomenon of dispossession of decision-making.
SM: There is a shortage of personnel in education. So technology becomes the panacea. Running out of premises? Distance learning. Short of speech therapists? The AI will make diagnoses. We are replacing human functions with a machine. Now, these children are entrusted to us and we must take care of them!
SM: Maybe, if it’s one activity among others. But do we want to replace school with that? That’s what we’re looking for. There is a clearly expressed desire, particularly at the OECD, for schools to switch completely to this, to eventually replace human transmission. Basically, what we are saying is that we could perfectly do without teachers. The problem is there.
EM: Hannah Arendt was already saying in 1961 that the gamification of education was a problem. It’s got to be fun… when the balance sheet of liabilities is so big!
The candy side of gamification makes us forget all the negative aspects. We always say that young people want fun, and therefore that it takes technique. It is the argument of legitimation which serves to bring in all this hardware. It is extremely worrying, what is happening, and there is a candor around it that is not up to the drama that is being played out.
EM: We are asking for a moratorium. We think we should stop going any further in this direction, before there has been a serious reflection on the problematic aspects that are raised by many specialists. Is this really what we want for our children? There is no democratic debate on this and we are pushing forward, on the pretext that we are not stopping change. And we never listen to teachers!
SM: We’re seeing it right now with the ChatGPT debate. We should say that ChatGPT must adapt to the school, when it is rather the opposite that happens: it is the school that adapts to ChatGPT!
EM: We get sent guides on how to integrate ChatGPT to make our lesson plans. From the moment we do that, let the robots give lessons for us. Me, I did 12 years of study, 10 years of teaching, if a machine does the lesson plan for me, where are we?
SM: My first instinct would be to tell you that I have a lot of compassion for secondary and primary teachers, who work in terrible conditions, and I understand that any help is welcome. If the working conditions were better, we might not need to resort to this. It is sad to rely on an algorithmic process to generate a creation that is an integral part of education.
EM: The profession of teacher is built on the mastery of knowledge and, then, the transmission of this autonomy. We now ask the machines to do it for us! We have to be very careful not to let the machines divest us of our autonomy. If a doctor consults Google to find information, no worries. But from the moment he lets the machine do the diagnosis for him, we have a problem.