To give you another perspective:
Let us assume that the world had only one risky/noisy asset $S(t)$ and let us further assume that at time $T$ our process cann only have $n$ states - namely $(S_1, \dots, S_n)$ and that the interest rate was flat and given by $r$
Now let's say we have a payoff funtion $f(x): \mathbb{R}\to\mathbb{R}$.
Working under the risk neutral measure $Q$ the time $t$ price of the derivative paying $f(S(T))$ at time $t=0$ is given by
$$
V(0)=e^{-rT}\mathbb{E}^Q[f(S(T))]
$$
Now we know that $S(T)$ only has $n$ different states and can thus decompose above expectation into
$$
V(0)=e^{-rT}\mathbb{E}^Q[f(S(T))]=e^{-rT}\sum_{i=1} \mathbb{E}^Q[f(S_i)]
$$
Thus our price is determined by the expected present values of the
different cash-flows that can be generated by our instrument/product.
In above case one would actually already know the price for every function $g(x):\to\mathbb{R}\to\mathbb{R}$ if all the probabilities
$P_i=\mathbb{P}^Q(S(T)=S_i)$ were known.
The price would then be given by
$$
V(0)=e^{-rT}\mathbb{E}^Q[g(S(T))]=e^{-rT}\sum_{i=1} P_ig(S_i)
$$