Full Idea
It seems unlikely the psychological process could mirror Lewis's semantics: people don't imagine a multiplicity of worlds and the pick out the most similar. Rather we construct representations of possible worlds from counterfactual descriptions.
Gist of Idea
We don't pick a similar world from many - we construct one possibility from the description
Source
Tim Maudlin (The Metaphysics within Physics [2007], 1.5)
Book Reference
Maudlin,Tim: 'The Metaphysics within Physics' [OUP 2007], p.33
A Reaction
I approve of fitting such theories into a psychology, but this may be unfair to Lewis, who aims for a logical model, not an account of how we actually approach the problem.