🧪 AI Advanced AI Teacher Simulation

An AI teacher does two jobs at once: it estimates how well a student has actually learned something, and it decides what happens next — a pass, a harder lesson, an intervention. This simulator makes both jobs visible and testable. Two groups of simulated students are plotted in 3D by true ability against the AI's predicted score; a systematic bias slider skews one group's predictions away from the truth exactly the way real deployed grading and tutoring models have been shown to, a debiasing-correction slider lets you claw that bias back out, and a threshold slider decides who passes. Live readouts track the demographic parity gap between the two groups, the calibration error of the corrected scores against ground truth, and the overall pass rate — so you can see, quantitatively, the ethical trade-off at the center of personalized learning: a model can look accurate on average while still treating one group of students unfairly, and fixing that requires measuring fairness and accuracy as two separate numbers, not one.