At the Prediction, the uncertainty
The state vector is updated by weighted averaging.
The state covariance
The Kalman gain
the state estimate is drawn towards the current measurement,
the uncertainty in the state estimate is reduced.
See the 1D case.