Tech • AI • Robotics • Game
ChatGPT vs School: The Great Danger for Our Children?
5/10TL;DR
The spread of AI in higher education is reshaping learning faster than universities are adapting, with evidence of both easier access to help and weaker independent mastery.
KEY POINTS
Grades surge at home, collapse in proctored exams
At Brown University, an economics professor reportedly saw average marks on take-home work climb to 96 out of 100, then fall to 48 when students sat a final exam in class without AI assistance. The comparison suggests that polished assignments can hide major gaps in comprehension. Similar concerns are surfacing in other elite institutions, where strong written submissions are not matched by oral explanation or in-room performance.
Students are using AI almost universally
In France, an Ipsos survey for EPITA found that 94% of students have already used AI, with 48% saying they use it every day. Nearly one in two said they had used it even when rules explicitly banned it. The survey also found that many students now struggle to imagine studying without these tools, especially in scientific and engineering tracks.
Professors are losing direct contact with students
Some faculty members in the United States report that office hours are becoming quieter because students now turn first to ChatGPT, Claude or similar tools. AI offers instant answers, no social pressure and constant availability. That convenience may reduce anxiety for some students, but it also weakens the human exchange that universities traditionally rely on to build confidence, judgment and intellectual maturity.
The problem may be pedagogy as much as technology
Traditional homework and essay-based grading are proving easy to outsource. As a result, some instructors are dropping take-home assignments, moving toward in-person exams, oral defenses and project-based evaluation. The shift reflects a broader conclusion: if assessment only asks for a standard answer from point A to point B, students will use the fastest available tool to produce it.
AI use does not equal AI mastery
Heavy use of chatbots does not mean students understand AI deeply. Many can ask a question and copy the answer, but far fewer know how to structure a workflow, connect tools, manage projects or test an output against real-world constraints. The gap between casual use and actual competence is becoming a central issue for employers and educators alike.
The missing skill is not prompting but initiative
Several educators argue that the key value now lies in what AI does not supply on its own: initiative, judgment and decision-making. If AI can do 80% of a task, the remaining 20% becomes more important, not less. That last stretch includes defining the problem, spotting errors, adding originality and deciding what is worth building in the first place.
Critical thinking remains weakly defined but crucial
Teachers increasingly want students to challenge AI outputs instead of treating them as finished truth. In practice, that means checking sources, testing claims, identifying blind spots and arguing back when the answer is shallow or wrong. The risk is that students use AI to avoid friction, when intellectual growth often depends on exactly that friction.
Different countries are taking opposite approaches
France has moved toward treating unauthorized AI use in assignments as a form of cheating. China has taken a more systematic route, introducing AI exposure from the age of 6 while maintaining strict limits during high-stakes exams. The contrast highlights two competing models: prohibition in assessment versus structured integration into learning.
Teaching methods may need to change radically
Long lectures and passive listening appear increasingly ill-suited to students accustomed to interactive digital environments. Some educators say lessons must be broken into shorter segments with quizzes, oral exchanges, timed exercises and active participation. The issue is not entertainment for its own sake, but sustaining attention and turning students from spectators into participants.
AI can also work as a tutor
Used differently, AI can reinforce learning instead of replacing it. Customized tutoring systems can quiz students on their own course material, adapt tone and examples to their interests, and provide repeated practice without judgment. In that model, AI acts less as an answer machine and more as a guided coach, helping students work through difficulty rather than bypass it.
CONCLUSION
AI is not simply making students smarter or less capable; it is exposing how outdated many teaching and assessment methods have become. The central challenge is no longer whether students will use AI, but whether schools can redesign learning so that understanding, curiosity and judgment still matter.
Ask a question