In AI projection can be done by running the progam, provided the complete state of the system is available, and the program describing the evolution of the system is available.
Suppose the neither a complete description of the situation nor a complete description of the process is available. Situation calculus ([McCarthy and Hayes, 1969]) was invented to deal with this by allowing logical inference from facts about a situation for which we do not have a complete description.
However, it is possible that humans do some projection by non-logical
means, i.e. by running some mental process on a representation of the
situation. If so, such mental simulation must use a greatly
schematized representation of situations. The idea is to be able to
answer a question, ``What will happen if ?'' by mental
simulation. A program could also use some kind of schematized
simulation. Its output seems to be propositional. Subjectively, such
simulations usually seem to be fragmentary.