When I was a very little girl I imagined that there were tiny people in the radio. I was very glad to get a different explanation from my parents. Tiny people inside things bothered me then and they bother me now.
Chapter 1, Who is the Controller of Controlled Processes? in The New Unconscious is written by Daniel Wegner. Here is his introduction to The Homunculus Problem
In the sciences of the mind..(the term, homunculus)..is pejorative It stands for an absurd explanation an inner executive agent who does the persons actions. Freuds theory of id, ego and superego, for instance, has often been criticized as a homunculus-based explanatory system in which the persons behaviour is explained by reference to an inner agent (in this case, a committee of them) that is responsible for the persons actions. Whenever we explain a behaviour by saying that some person like agent inside the person simply caused it, we have imagined a homunculus and have thereby committed a classic error of psychological explanation
The issue here, of course, is that a homunculus must itself be explained. The path of explanation implied by the homunculus idea is to reapply the same trick and suggest that another smaller homunculus might be lurking inside the first. This path leads to the specter of an infinite regress of homunculi, nested like Russian dolls. That quickly descends into absurdity. Another way to explain a homunculus is simply to say that it has free will and can determine its own behaviour. This means the homunculus causes things merely by deciding, without any prior causes leading to these decisions, and thus renders it an explanatory entity of the first order. Such an explanatory entity may explain lots of things, but nothing explains it. This is the same kind of explanation as saying that God has caused an event. A first-order explanation is a stopper that trumps any other explanation, but that still may not explain anything at all in a predictive sense. Just as we cannot tell what God is going to do, we cannot predict what a free-willing homunculus is likely to do either. There cannot be a science of this.
Most psychologists and philosophers are well aware of the homunculus problem and it has been generally avoided in contemporary theorizing, with one noteworthy exception. The notions of controlled and automatic (brain) processes carry with them the implicit assumption of a kind of homunculus. Now it is true, of course, that most current cognitive and social cognitive research focuses specifically on the automatic side of this dichotomy, so much so that there seems to be progressively less room for the little person in the head. But why should there be any room at all?