Resolving uncertainty on the fly: Modeling adaptive driving behavior as active inference