When AI Assists Students, Who Is Actually Thinking?
Several years ago, when students were given an essay assignment, the first challenge they usually faced was finding material and understanding what they actually wanted to write. They opened books, searched for articles, read several sources, and then tried to construct an argument, even if the process did not always go smoothly. Some misunderstood the readings, struggled to determine their position, repeatedly deleted paragraphs, or only discovered their main idea after writing at length. Such a process was indeed time-consuming, but it was through this process that the ability to read, weigh information, and formulate thoughts was slowly developed.
The arrival of generative artificial intelligence has changed this process rapidly. With a few command sentences, students can now obtain an essay framework, article summaries, alternative arguments, and even a neatly structured draft essay. Tasks that previously took hours can be completed in minutes, and this development certainly presents many opportunities for higher education.
It is difficult to imagine campuses distancing themselves from AI. Students are already using it to understand concepts, translate readings, seek alternative explanations, improve their writing, and even discuss issues. Lecturers are also beginning to utilise the same technology for various academic purposes. The debate over whether AI should be accepted or banned in universities is therefore increasingly losing relevance. The technology is already in the study room, and its use is likely to become even more widespread.
A more fundamental question arises when AI begins to take over roles in almost every stage of a student’s work. If a machine helps search for information, summarise readings, construct arguments, write paragraphs, and draw conclusions, to what extent is the student still undergoing the thinking process that is supposed to be central to higher education? This question is important because successfully completing an assignment does not necessarily indicate that quality learning has taken place.
Higher education has never been solely concerned with the final product, be it a paper, report, presentation, or thesis. Behind that product lies a more important intellectual process, starting from understanding the problem, reading critically, comparing viewpoints, connecting information, testing arguments, to justifying a conclusion. Students do need to produce good writing, but the quality of education certainly cannot be judged merely by how neat the papers they submit are.
This is where the ability to think critically becomes important. Peter Facione, through his classic study on critical thinking, places skills such as interpretation, analysis, evaluation, inference, explanation, and self-regulation as part of the critical thinking process. In simpler terms, it is not enough for a student just to know an answer. They need to understand the problem, assess the available evidence, consider alternative explanations, test the consistency of an argument, and be able to explain why they ultimately reached a particular conclusion. Such abilities are not formed merely by knowing their definitions, but through repeated practice.
The issue becomes intriguing when linked to a principle in neuroscience, namely ‘use it or lose it’. Jeffrey Kleim and Theresa Jones discuss this principle in the context of experience-dependent neural plasticity, which describes how experience and the use of certain functions relate to the alteration and maintenance of neural function. They also introduced the principle of ‘use it and improve it’, which emphasises the importance of practice in strengthening a capability.
This principle cannot be carelessly translated into an assumption that using AI will damage students’ brains. Kleim and Jones’s research developed primarily in the context of neurological rehabilitation, not the use of AI in learning. However, the basic idea raises a relevant question for education: if abilities develop through use and practice, what happens when the opportunity to practice those abilities is increasingly handed over to technology?
Cognitive psychology offers another concept that helps explain this issue, namely cognitive offloading. Evan Risko and Sam Gilbert explain how humans use external sources to reduce the burden of cognitive work. We do this every day, for instance by storing phone numbers on digital devices, entrusting schedules to calendars, and relying on calculators for certain computations. Technology is indeed created to help humans work more efficiently.
In education, however, the issue is slightly different because some of the work that feels difficult is actually part of intellectual training. A student who uses AI to get an additional explanation of a term they do not yet understand is certainly different from a student who asks AI to read sources, compare arguments, determine a position, and then compose the entire paper. In the first situation, AI can help the student understand the problem. In the second situation, the student may get the work done faster, but the intellectual process they should have undergone becomes increasingly minimal.
We often advise students to double-check the answers produced by AI. This advice makes sense, but the ability to check an answer also requires knowledge. How can someone recognise an error in AI’s output if they themselves do not yet sufficiently understand the issue being discussed?