We go back to university and we haven't done our homework. We're a year into generative AI, and unlike our students, we still don't know how to use it in the classroom. We have read, written, debated and watched TED talks, but the reality is that we are all a bit Manel Ferrando, that character from Persones Humanes who, when faced with any topic, always said “I don't understand too much, but I have read a lot about the subject.” ”.
In the academy there is concern and it is measurable. At UPF, my campus colleagues have made a series of highly recommended critical articles on the subject. Professor Ethan Mollick of the University of Pennsylvania, a leader in the application of AI in the classroom, directly describes it as chaos. He affirms that once we have overcome level 1 of panic – written assignments are no longer useful for evaluation – we no longer know what to do.
We are used to a 10-15 year emergence cycle of new technologies. In the mid-1960s, companies became computerized, in 1981 the first IBM PC arrived, in the early nineties the web, in 2007 the iPhone, the rise of social networks and, now, AI. So far predictable. What was not is the adoption rate. All of these “thinking tools” (in the words of philosopher Daniel Dennett) usually take about five years before the social impact is seen. The adoption of AI in classrooms has been immediate and the revolution has come from below.
The immediate use case for ChatGPT is to generate academic papers; from the commentary on a book by Rodoreda to the causes of the Great War, including how to explain Brownian motion to ten-year-old children. Writing, especially when you are a student, is a heavy and repetitive task and wasn't this type of task that technology should free us from? Green sticker for students in technology application.
Since the emergence of Google and Wikipedia, written text as a reliable evaluation method was in crisis. Websites like Patatabrava.com and online text remixing services had already forced academic institutions to invest in plagiarism detection software and teachers to dedicate part of their time to acting as a watchman. However, these types of tools do not detect texts generated on request by companies that are dedicated to this or texts generated by platforms such as ChatGPT. The young generative AI is just the latest iteration in this game of thieves and watchmen; He is the boy in Andersen's story who tells us that the king is naked.
The sooner we admit that we are pretty slow when it comes to AI in the classroom, the sooner we will know how to dress for when the cold weather arrives. The debate is not whether or not generative AI should enter the classroom; The debate is what we do and how we do it. I have some idea.
To begin with, we should learn from the fiasco that the digitalization of primary and secondary classrooms has entailed. Suddenly we realized that we ran too much for fear of being classified as technophobes and we digitized in droves. One of the most disastrous consequences has been the supremacism of PowerPoint. Teachers who reduce their classes to reading a series of points on slides, which is the same thing students will do at home. All this while the subject book waits closed at the end of the course and Wallapop.
The other big mistake is having broken the brain-hand loop when taking notes. On the keyboard we can write everything the teacher says without having to understand anything. If, on the other hand, we take notes by hand, we will need an extra effort of understanding and synthesis before transferring it to paper; by hand we are not going so fast. Another of the big problems of digital media is reading comprehension. Independent studies show how students who read on paper achieve better reading comprehension scores than those who read the same text on screen.
But in the same way that generative AI could do away with written assessment of work, it could also be the Holy Grail of the classroom in a post-industrial society. Years ago we realized that in an environment of infinite information and digital resources it makes little sense for us to still build the learning process around the master lesson, a model inherited from the industrial revolution. Today we are asked to reverse roles in what is known as the flipped classroom model, a flipped classroom where students learn at home, with books, articles, movies, YouTube channels, tiktokers and AI, and share everything they learn online. Classroom. The faculty recovers its role as a space for experiences and horizontal learning (yes, going to play botifarra at the bar counts) and the digital space does what it does well: transport information. Reversing roles is a guarantee of failure, not only academically, but also socially.
In this sense, in the subject that I teach on audiovisual technology at the UPF, seven years ago, to lead by example and with the support of Amical Wikimedia, we made a wikiproject for students to contribute what they learned in the classroom . If until now you have copied from Wikipedia, now you have to do the opposite by sharing your knowledge in the form of contributions that are evaluated. In a way we have flipped the web.
Could we do the same with text generation tools and also reverse their use? Well we could try it. If we accept that generative AI has all the answers even if they are not correct, the merit will be in knowing how to ask the right questions. So, perhaps what we should start doing is not evaluating the answers but the questions, the references, the critical vision, the method, the dialogues they generate, the paths they choose and the personal contributions. In fact, asking the right questions and doubting everything is the basis of the scientific method, one of the values of academia.
The philosopher Rubert de Ventós defined philosophy as “the restless ass of the spirit” and the philosopher as “he who sees blurry where everyone sees clearly.” If we turn it around, AI can be the glasses that help us see blurry.
