My criticism can be summarized as "map is not the teritory".
The reality is messy. Sure, we can think in terms of feedback loops and interactive modules and it can be helpful sometimes. We just have to be aware we are imposing our puny models onto the infinitely complex beast.
The model will break down at some point. Recognizing the correct model and extrapolating from it is not useful unless we know the limit at which it breaks down.
It is also hard to be objective about our mental models, we tend to apply all kinds of favoritisms and observation biases when evaluating them. In machine learning (google maps) those biases are still present, but at least they are not our personal biases so they are easier to spot and correct for.
I'm arguing for "most things can in most cases be considered through lens of simpler system" instead. "Everything is a system" to me implies that every phenomenon maps to exactly one comprehensible system, which is far too strong statement to make.
At any moment, the employed model used to understand a given situation is local. The model is adaptive in the sense that you only need to concern yourself with what is immediate, and concerns about how the effects might propagate are limited to the interface between the immediate moment and the external ones.
Also, many models can be used simultaneously and overlap in the analysis of a particular situation. It isn't necessary to Have It All Figured Out and keep it all in memory simultaneously, just to have a set of methods to draw quick conclusions. Ensemble methods tend to work best IME.
> You can say the right words to make a date happy.
If the partner doesn't like you, you can say all the right, do all the right things, etc and still be passed-over.
Then another partner comes in and does all the wrong things. Still he gets not passed-over.
There's chemistry involved in relationships as well. From smell to much more subtle things. It's a function with infinite variables. We can create models that help us understand society, economy, etc. But these are social sciences, the outcome is affected by the observer ;-)
