Practical AI hosts say schools lag AI-era work, citing study where an unrestricted GPT-4 chatbot cut later exam scores 17%

How a tutor is built around the model mattered more than the model itself. In a peer-reviewed field experiment with nearly 1,000 Turkish high-school math students, open GPT-4 access raised practice scores by 48% but lowered later no-AI exam scores by 17%. A tutor version that gave hints instead of answers avoided that harm. A separate randomized crossover trial with 194 Harvard physics students used a structured GPT-4 tutor built on expert-written prompts. It produced more than double the learning gains of an active-learning class across two lessons. Both studies show that scores measured while students have AI can mislead, and that a later test without AI is what reveals real learning. [3] [5] [4]

The hosts argue schools should treat AI as a permanent tool while keeping the useful struggle that builds critical thinking. Their alternatives to banning AI and running detection tools include having students critique AI-written essays and letting teachers experiment. Teachers have pushed back that classroom policy is set above them, so the hosts say change has to come through districts and state requirements. [2]

Purdue has made AI competency a requirement for every undergraduate degree. Trustees approved it in December 2025, and it starts with students entering in fall 2026, who can meet it through 22 courses of 1–4 credits each. Critics argue that relying on AI wears down core skills. [6] [8] [9] [7]

Whitenack says daily use of AI assistants is a basic expectation at his company, Prediction Guard. He worries that university AI-detection policies leave graduates reluctant to use AI at work. This is his own hiring and mentoring experience, not survey data. [2]