While I agree with the overall sentiment, I think it’s important not to overdo this approach. Let me explain.
Consider the situation where you have a stochastic process which generates values—for example, you’re drawing random values from a certain distribution. So you draw a number and let’s say it is 17.
On the one hand you did draw 17 -- that number is “real” and the rest of the distribution which didn’t get realized is only “imaginary”. You should care about that 17 and not about what did not happen.
On the other hand, if we’re interested not just in a single sample, but in the whole process and the distribution underlying it, that number 17 is almost irrelevant. We want to understand the entire distribution and that involves parts which did not get realized but had potential to be realized. We care about them because they inform our understanding of what might happen if the process runs again and generates another value.
Similarly, if you treat history as a sequence of one-off events, you should pay attention only to what actually happened and ignore what did not. But if you want to see history as a set of long-term processes which generate many events, you’re probably interested in estimating the entire shape of these processes and that includes “invisible” parts which did not actually happen but could have happened.
There are obvious methodological pitfalls here and I would recommend wielding Occam’s Razor with abandon, but that should not conceal the underlying epistemic point that what did not happen could be important, too.
While I agree with the overall sentiment, I think it’s important not to overdo this approach. Let me explain.
Consider the situation where you have a stochastic process which generates values—for example, you’re drawing random values from a certain distribution. So you draw a number and let’s say it is 17.
On the one hand you did draw 17 -- that number is “real” and the rest of the distribution which didn’t get realized is only “imaginary”. You should care about that 17 and not about what did not happen.
On the other hand, if we’re interested not just in a single sample, but in the whole process and the distribution underlying it, that number 17 is almost irrelevant. We want to understand the entire distribution and that involves parts which did not get realized but had potential to be realized. We care about them because they inform our understanding of what might happen if the process runs again and generates another value.
Similarly, if you treat history as a sequence of one-off events, you should pay attention only to what actually happened and ignore what did not. But if you want to see history as a set of long-term processes which generate many events, you’re probably interested in estimating the entire shape of these processes and that includes “invisible” parts which did not actually happen but could have happened.
There are obvious methodological pitfalls here and I would recommend wielding Occam’s Razor with abandon, but that should not conceal the underlying epistemic point that what did not happen could be important, too.
You make a good point.