Artificial intelligence arouses fears and hopes. The automation of certain tasks could eliminate 300 million white-collar jobs in the coming decades. Law seems to be one of the professional fields that could be the most affected. At the same time, new analytical capacities are already offering new ways to tackle certain social problems, such as school dropout. Welcome to your near future…
In the great class of online digital learning platforms, Moodle is a model student. About 80% of universities and CEGEPs in Quebec use it, like tens of thousands of other establishments around the world.
Hundreds of millions of teachers and students use it daily to organize lessons, transmit documents, plan readings and assignments, send personal messages or simply check activity participation. The pandemic has only accentuated the great “moiddleization” of the education planet. Moodle is the jackpot.
The Himalayas of data generated by this universal tool are now used to analyze learning through surveys using artificial intelligence (AI). Two weeks ago, at the University of Texas at Arlington, education researchers attending the Learning Analytics and Knowledge Conference (LAK 23) presented their ongoing work in this cutting-edge field.
Quebec professor Bruno Poellhuber was one of them. He unveiled the progress of his research in predictive analysis to report significant delays in learning, even the early detection of students at risk of dropping out.
“We want to try to predict school dropout,” summarizes Professor Poellhuber, speaking of the research entitled Smart dashboards and success aacademic (TIR-IA), co-piloted by Professor Lyse Langlois. “The real idea behind this project is to identify cases that pose challenges and facilitate corrective interventions. »
Refinement of the tool continues to analyze data generated on Moodle by undergraduate students at the University of Montreal (UdeM), in the fall of 2020. AI capabilities, including artificial neural networks and decision trees allow us to go much further than traditional statistical analysis. The model under development filters the data and gives it meaning by analyzing it.
“We worked on a methodology to extract the traces and format them to allow us to understand how the behavior of students – the items on which they click, for example – can help us predict dropout. We had some success. Our model performs better than chance, but it is not yet perfected. There is still a lot of work to be done to use it as a predictive tool. »
The Dashboard already shows explicitly how technology can improve feedback — from learners to teacher — perhaps the most fundamental of pedagogical steps: once the material has been transmitted, it is necessary to quickly check who retains it, and how . “Digital gives a lot of hope to enrich the feedback, sums up the UdeM professor. I’m one of those who see digital things in a rather positive light, although I remain nuanced. »
The era of robots
So it’s not just conversational robots that are already transforming education, even if this tool symbolically concentrates the power, the hopes and the fears of the revolution in progress. The Web as the e-mail box of the Duty abound with confidences from teachers or professors like Martin Dubreuil, who can be read today in our Opinion section, who have used or banned chatbots in class. Institutions (such as UdeM) have adopted rules to simply prohibit the use of conversational robots, considered plagiarism, “in the context of an evaluation, unless explicitly provided for in the evaluation instructions” .
“In the uses and uses of conversational robots, it’s rather emerging,” says Professor Normand Roy of UdeM. There are plenty of anecdotal testimonials everywhere, but no great studies on the actual uses of ChatGPT. We stay on the surface, to explain that such a teacher has tried such a thing or that at Laval University, a teacher authorizes its use in his courses. »
Professor Roy participates in the research of his colleague Bruno Poellhuber and leads the Interuniversity Research Group on the Pedagogical Integration of Information and Communication Technologies (GRIIPTIC), which brings together some twenty Anglophone and Francophone researchers. He teaches educational technologies and prepares a thematic issue of the journal Living primary on the digital skills to be transmitted. He and his research students are particularly interested in virtual reality, video games, digital laboratories and AI, of course.
“Before last September, we mainly talked about the ethics of AI, he says. We had no examples of concrete applications. Since the public launch of chatbots, the conversation revolves around these real applications. Elementary and secondary school teachers are forced to question AI literacy, what they really know about it. »
A leap of scale
Simon Collin holds the Canada Research Chair in Digital Equity in Education at UQAM. His research focuses specifically on various issues of equity or standardization of pedagogical practices related to the use of technologies in the classroom, including AI. ” In the endall of this aims to ensure that the use of technologies in education supports the objective of school democratization and does not hinder it, ”he summarizes.
He tried chatbots as soon as they appeared, was “amazed like everyone else by the capabilities available”. He speaks squarely of “a leap of scale”, not to say of revolution, and he quickly wondered how this new tool was going to transform his world.
“You have to find a pedagogical meaning in technical tools that were not necessarily originally designed for this educational use. Tools that have specific rather than very broad functions seem more promising. »
He gives the example of a software that allows you to ask questions to a historical figure, while ChatGPT has a much more extensive use. He adds that he has already adapted the course he is giving this master’s session. The exam questions are based on very specific content seen and discussed in class, relating to technological issues. A chatbot can’t answer it, according to the professor.
“I didn’t find any traces in the copies of ChatGPT,” says the professor, adding that this use would have been possible given the maturity of the graduate students. “For primary school students to baccalaureate students, the school aims to develop skills. We must therefore ask to what extent the performance of the tool supports or interferes with the development of skills. If you develop spelling or syntactic skills in primary school, for example, you should not use ChatGPT, which is so powerful. On the other hand, at the end of primary or secondary school, we may be able to use the chatbot to give leads and help develop creativity in writing. »
Above all, Professor Collin comes back to this idea that education remains an eminently relational approach. Artificial intelligence, like other tools, must therefore be used to support these relational activities and can even give them even more value. “Lack of time is one of the main obstacles to educational activities, and AI could, for example, save teachers time on administrative tasks to invest it in relations with the student. »
Precisely, Bruno Poellhuber is associated with a new group of a hundred professors who have just submitted a request for support from donors to create a “living laboratory on the learning of the future in higher education”. “Too much teaching is still done 100% with lectures and assessments with objective questions,” he says. More is needed. But you also have to make room for other proven methods. »