Home▸Machine Learning & Neural Networks▸Learning Metacognition Simulation

Metacognition & Confidence Calibration Simulator

Interactive metacognition simulator: run trials where a simulated learner reports a confidence rating alongside each answer, and watch a live 3D calibration curve reveal over- or under-confidence, self-insight, and Brier score in real time.

Machine Learning & Neural Networks2DModerate60 FPS📱 Mobile-adapted

Metacognition — awareness and control over one's own thinking — shows up most clearly in how well confidence tracks accuracy. This simulator runs a stream of trials for a simulated learner whose correctness and self-reported confidence both derive, in part, from the same underlying evidence. Three sliders let you dial in the learner's true task ability, their metacognitive insight (how much confidence is genuinely informed rather than noisy), and a systematic confidence bias. Every trial appears as a colored point in a live 3D calibration chart, while purple bars track measured accuracy per confidence bin against a white "perfect calibration" reference line — with running readouts for accuracy, mean confidence, overconfidence, and Brier score, the standard scoring rule for calibration quality.

⚙ Under the hood

This simulation illustrates the concept of metacognition – thinking about one's own thinking. It shows how awareness and control over cognitive processes can enhance learning outcomes.

Metacognition Simulation

2D · HTML5 Canvas 2D · 60 FPS target · runs fully client-side, no install