Synthetic Populations

Synthetic populations are useful when the real object of interest is not a single response, but a distribution of responses under changing conditions.

One simple way to write the simulation state is:

$$S_{t+1}=f(S_t,E_t,G)$$

where $S_t$ is the population state at time $t$, $E_t$ is the event or information environment, and $G$ is the social graph that controls who influences whom.

The interesting part is not the formula. It is whether the assumptions behind $f$ are visible enough to inspect, criticize, and improve.