[klabb3]

Exact numbers are easy to measure but not the only way to measure something. Data-driven is the new statistics but 10x worse because it has more sources of obfuscation, and bean counters (including super “smart” and analytically minded) obsess narrowly over models. Add ML/AI and you got an order of magnitude again.

There’s an impedance mismatch where we have enormous amounts of useless data and a small amount of useful data. Best we can do is use and create more of the useful data instead of obsessing over P>.99 on something useless that will be misinterpreted anyway.

The problem, in my view, isn’t qualitative per se, but rather unfalsifiability. Ideology is when the solution is always more of the same no matter what the outcome is. Communism/socialism and neoliberalism all fall in this category. I believe this holds true if you go more academic into Keynesianism and say Chicago school - models that have become truisms to their followers.