Drift-diffusion: how your brain trades speed for accuracy

本站添加: 2026-06-02

[00:00:00]Imagine you're making a decision between two choices. Each moment a little bit of noisy evidence trickles in. You add it up. When the total crosses a threshold, you commit.

[00:00:11]The math is simple. The change in evidence equals a steady drift plus random noise. Drift is the true signal.

[00:00:18]Noise is everything else.

[00:00:21]Now, run a hundred of these. Same drift, same noise, different random fluctuations. Most hit the correct threshold on top. A few unlucky ones cross the wrong one. Those are the errors.

[00:00:32]Collect when each trajectory finished and you get the reaction time distribution. Notice the long tail. Slow decisions are rarer, but they happen.

[00:00:41]Now, the speed-accuracy tradeoff.

[00:00:44]Here's the key insight. Lower the threshold and you decide quickly, but more errors slip through. Raise the threshold and accuracy climbs, but every decision takes longer. Speed and accuracy traded against each other by a single parameter.

[00:00:59]Plot accuracy against reaction time as you sweep the threshold and you get this curve. Every organism, every decision.