> I wonder if 'fuzzy' or uncertain models like economic ones can tolerate or benefit from slightly less deterministic algorithms. It depends on what sort of fu

> I wonder if 'fuzzy' or uncertain models like economic ones can tolerate or benefit from slightly less deterministic algorithms. It depends on what sort of fuzziness or uncertainty you're talking about. If you're referring to a model's sensitivity to initial conditions / measurement noise, then a common solution is to use Monte Carlo simulation, in which case you get an estimate of a probability distribution function as an output (which is often collapsed to an expected value, but that's not necessary). I'm not aware of any analog computers that give probability distribution functions as outputs rather than single values.