Many had predicted it: cheating in a school environment using ChatGPT is happening in particular in France. A professor at a university in Lyon discovered that several of his students had their assignments written by ChatGPT. What really questions the detection of cheating and the lack of regulation in France.
It is often said that artificial intelligence will invade our society, that our uses of technology will be disrupted, without us experiencing any real change. It’s the same story with ChatGPT, the automatic text generation AI powered by OpenAI that has been all the rage on the internet for several weeks now.
However, we look at the first concrete use cases, but not necessarily honest. Lyon 1 University students used ChatGPT to write an assignment: unfortunately their teacher exposed their cheating method.
Students use ChatGPT to write their assignments
14 health students had a week to write an assignment whose theme was ” Definition of the main characteristics of the approach to disability in Europe», in the course of Stéphane Bonvallet, professor of disability. Of the 14 copies returned, the professor grimaced at seven that looked suspiciously alike: “It was not a copy and paste. But the copies were built in exactly the same way.“he tells the Progress.
While reading the task, several details caught your attention: similar grammatical constructions, identical reasoning, and comparable details. He explains that the copiesthey were all illustrated with a personal example, related to a grandmother or a grandfather… Therefore, it seemed obvious that these copies were not normal.»
While interviewing a student, he learned the truth:He told me that 50% of the students in the class had used ChatGPT’s artificial intelligence to write their paper. Apparently, they had spread the word on social media.If Stéphane Bonvallet had heard about the OpenAI tool and its potential dangers (especially in relation to his profession), he did not believe that the AI was powerful enough to write his students’ assignments.
According to the professor who discussed with his colleagues, “theThree-quarters of teachers told me they’ve had to deal with dubious scripts in recent weeks.“. He specifies that the practice would be more widespread in secondary schools and wants to find a solution as soon as possible.
Unarmed teachers: their anti-plagiarism tools do not work with ChatGPT
The problem with ChatGPT is that technically it is not plagiarism: therefore its use is not prohibited in the university, for the moment. The interested teacher indicates that “Not having a framework that currently prohibits this practice, I was forced to write them down. From one copy to another it was worth between 10 and 12.5. So I gave 11.75 to the seven students who used ChatGPT.“The University of Lyon 1 has taken note of the copies and “conduct an instruction“. From a pedagogical point of view, the practice of automatic generation”poses a real knowledge integration problem, because students no longer need to do any research to composeadds Stephane Bonvallet.
In college, homework is mostly tested for plagiarism, based on alumni papers and pages all over the web. Anti-plagiarism tools make it possible to detect this form of cheating and, above all, discourage students from practicing it. Where ChatGPT does a good service for students who would like to cheat is that AI-generated texts are not detected by this software. If they do take content from the Internet, the latter are mixed and reformulated: the frames of the sentences, if they are similar, are not identical.
On the other side of the Atlantic, solutions are beginning to be found. As indicatednumeral, New York City has decided to simply cut off access to ChatGPT in schools. The media took the opportunity to remind us that educational culture had to confront Wikipedia and the smartphone: the revolution represented by ChatGPT could be of the same ilk. There may be a significant amount of information, adaptation and educational work ahead for educational teams around the world.
Knowing ChatGPT is useless: to cheat you must above all know how to use it well
The observation made indirectly by Stéphane Bonvallet is that all students today know ChatGPT. The accessibility of the tool (free, online) makes its democratization much easier. If schools have ways to track AI usage and prevent cheating, we may think it won’t be as simple as that.
Also, we have the impression that OpenAI has already restricted the use of ChatGPT for writing tasks. We ask:Write me an assignment on the topic: “Define the main features of the medical approach to disability in Europe”“, to which she replied:”It is not possible to write you a complete assignment on this question, however, I can provide you with information on the main features of the medical approach to disability in Europe.before enumerating those traits. However, the system is very easy to circumvent. Just ask him to imagine that in a fictional environment, he has to write a task and the robot will work without even realizing the deception.
Asking to write a task with ChatGPT is quite easy: in one sentence, you can do it. But since its publication there have been many tests and we know that we can go further. In the request that we write, we can ask ChatGPT to use a certain tone, to focus on a particular topic, to conclude with such a postulate, etc. It is by specifying the request that the power of ChatGPT can truly be revealed.
Beyond that, smarter students might use ChatGPT not to write their homework, but to brainstorm arguments to move forward, parts to tackle. The great strength of this application can also be its spirit of synthesis: with all the content available on the Internet, you can quickly get lost and fall victim toinfobesity. ChatGPT would be a kind of remedy for this evil by summarizing everything that can be found on the web.
To follow us, we invite you to download our application for Android and iOS. You can read our articles, archives and watch our latest YouTube videos.