Passive BCI (implicit mental-state monitoring)

In the taxonomy of Zander and Kothe, BCIs divide into active (the user issues a voluntary command), reactive (the system reads a response to an external stimulus, as in a P300 speller), and passive: the interface monitors spontaneous brain activity to infer a cognitive or affective state that the user is not deliberately producing, and adapts the system accordingly. No control is intended; the brain signal is used as an implicit channel about the user rather than a lever.

Typical targets are mental workload, drowsiness and vigilance, engagement, surprise, and error processing — the last exploiting the error-related potential, a signal the brain generates automatically when it detects a mistake, which a machine can use to veto or correct its own action. Passive BCI is attractive for the real world precisely because it asks nothing extra of the user, but it inherits the field's hardest evaluation problem: it is easy to decode a state that merely correlates with task difficulty or movement, and hard to show the decoder tracks the intended construct.