I have posted a similar question earlier and I guess this sounds naive to all of you, but nonetheless let me just ask:
Consider I have a simple and deterministic model $M$, with a number of input parameters. Given the fact that we do not have exact knowledge about the parameters, we can run $n$ simulations of $M$ with possible parameter values, say
Suppose for the sake of simplicity, we know that all input parameters are uniformly distributed with support [5,10].
Now, here is the question: What is the benefit of evaluation the model $M$ with $n$ randomly chosen parameter values (according to the uniform distribution with support [5,10]), rather than $n$ deterministically determined parameter values in the interval $[5,10]$, e.g. by assigning the values to $(5, 5+\frac{(10-5)}{n}, 5+\frac{2(10-5)}{n}, 5+\frac{2(10-5)}{n}, \dots, 10)$?
Consider that I run the model exactly $n$ times.