Because MVC not [only] models the user's mental model, it models the world.
It's nothing else, than a practical implementation of Hegelian dialectic.
Let's start with a simple scheme. The user sits besides a computer. [S]he looks at the monitor, while trying to achieve something by pressing buttons on the keyboard, or trying to wander around with a mouse or on a touchpad.
Generally, we don't really distinguish between the different input peripherials with MVC. The monitor is the output, other human-computer interface peripherials are for input, it's that easy:

(Based on image of flickr user ntr23 under Creative commons license)