Getting solutions while losing understanding – these are the biggest dangers of AI in education

Artificial intelligence could soon completely transform the framework of education, and this process will have a significant negative impact on the common way knowledge is acquired. This is what Corvinus teachers Márton Benedek and R. Balázs Sziklai see as the main threat of AI in education, as they have allowed their students to use tools like ChatGPT for two semesters without restriction in their mathematics courses. The two instructors had expected an increase in student motivation from allowing the use of AI, but they achieved the opposite effect:  

The good ones don’t win 

“Bad learners were happy with the easier way to solve problems, while good learners realised that they could achieve the same results with less effort, so they stopped bothering with traditional learning methods,”

says Márton Benedek, adding that the important phases of creating, making mistakes and understanding were lost in the learning process when solving problems with AI.  

In the spring of 2023 – the first time Corvinus faculty members took an AI-enabling course – ChatGPT had only been around for a few months, and only a few students were using the tool. But in their course a year later, almost all students were using ChatGPT-like platforms. Within a year, the use of AI had become so widespread and the tools so sophisticated that the majority of students had blindly trusted them without even trying to understand the process that led to the solution of the tasks.  

“Artificial intelligence has replaced the process of finding solutions to problems, when in the process of knowledge acquisition, understanding the process would have been much more important than the solution itself,”

explains Balázs R. Sziklai. He noted that most of the students who passed the exam with perfect scores using AI had no idea how they had performed between 0% and 100% until they received their scores. 

Preference for talented students 

Moreover, in addition to reducing the overall level of knowledge and motivation of students, intensive use of AI can also exacerbate social inequalities between students. Indeed, more affluent students who have access to more advanced subscription versions of ChatGPT-like tools can produce much better submissions than their peers who use the free versions.  

Despite their negative experiences, Márton Benedek and Balázs Sziklai R. do not see the solution in banning AI-based tools, but they consider it necessary to limit the use of AI-based tools in education:

“the ideal goal, in our opinion, would be for students to be able to use these tools while being aware of the basis on which ChatGPT produces a solution, and to be able to detect and correct errors made by AI in solving a task. We have found that if a student just pressed the START button and trusted ChatGPT completely, he or she could not assess the severity of the problem and was not sure whether ChatGPT was providing the correct solution,”

adds Balázs Sziklai.