Reductionism vs Systems Thinking
Two ways to break down the problem
Whether you’re in science, medicine, or engineering, you’re trying to understand systems.
A system is something that takes an input and gives an output. The system may have a lot of parts inside of it that are involved with transferring inputs into outputs. There’s a lot more nuance, but this is good for now.
There are two main ways to analyse a system — reductionism and systems frameworks. What exactly are these?
Here’s how I think of them.
Reductionism
Reductionism centers the parts of a system. It boils down to studying how each part relates to the output.
Doing this requires us to set up experiments where we fix all other parts, manipulate one part that we hypothesize relates to the output, then statistically determine whether we saw something.
The idea here is that we can sum up what we learn with each part<->output relationship and, viola, we understand the system.
Systems Frameworks.
Systems frameworks center the output(s) of the system. They boil down to inferring how each output arises from the dance of several parts.
Doing this is a lot more flexible — it can be done with experiments + high-dimensional statistics, or it can be done observationally with some epistemic uncertainty.
My favorite way is to bring in control theory and treat successful interventions and/or predictions in place of statistical tests. This is basically what happens in medicine, and it’s surprisingly effective given how vast the potential solution space is.
Summary
Reductionism tries to understand how a system generates an output by studying how each individual part relates to the output — then building up the whole from the parts.
Systems frameworks start with an output, or function, then finds out all the relationships that lead to it — then integrating multiple subsystems together in a way that respects interactions between them.