History only runs once, but inside a computer a virtual model of the historical past would run infinite times. In the computer, we would explore (by altering the variables) the entire possible range of outcomes for different past behaviors. The idea is then simulating inside a computer what we know about actions having been performed in the past and experimenting with the effects they may produce in such a virtual world.

That means that, inside a computer, the Past would be seen in the Present as a sequence of finite states of a temporal trajectory. Such a simulation would not “see” the past as it once was nut as potentialities for action, that is, explanations that can take place when it encounters a situation of some sort.

A computer simulation should allow us to understand archaeological observables in terms of a priori affordances: relationships between observed properties and the inferred properties/abilities of people having generated those properties. The affordances of any archaeological evidence become obvious in its use and/or formation process. Both involve establishing and exploiting constraints (between the user/producer and the material evidence of his/her action, the user/producer and the natural environment, and the material evidence and the natural environment).

In this paper we review the very idea of mechanical explanation defining a mechanism as organized collections of entities and activities that produce regular changes. Under such an account of mechanism, I think we can build a new way of explaining what a social science should be. In so doing, I totally agree with the so called "analytical sociology" approach: the goal of a social scientist is to explain an empirical phenomenon by referring to a set of entities and processes (agents, action and interaction) that are spatially and temporally organized in such a way that they regularly bring about the type of phenomenon the social scientist seeks to explain.

In any case, the resulting model is just an hypothesis. It may be more explanatory than the same hypothesis expressed in verbal terms, but it is not yet an explanation. Truth is what the world is, and an artificial society is out of the world. So, we cannot search for validations within a simulated model. Any model needs to be fed by, but at the same time provides feedback to, two theoretical complements which must be formulated independent of the simulation itself: (a) a bottom-up theory of agent interaction (local rules), and of the process from them to social organization (macroscopic effects); (b) a theory of downward causation, showing how agent interaction is modified by the differentiation it contributes to achieve.

In the paper we will consider more deeply two levels of analytical explanation:

• A theory that explains the knowledge about the state of the world an agent generates.
• A theory that explains the believes about social expectations an agent generates.