modeling
notes on mathematical and conceptual modeling. published on moonflowers.xyz.
the core idea
"all models are wrong, but some are useful." — George Box
models aren't truth. they're tools. the question isn't "is this model right?" but "is this model useful for what i'm trying to do?"
examples across domains
- anesthetics EEG model — a simplified representation of brain activity that's useful for monitoring patients, even though it doesn't capture everything happening in the brain
- data center environmental impact — models that estimate carbon footprint, useful for decision-making even if imprecise
- linear regression — the simplest useful model. often good enough.
- firefighting strategy — models of fire spread that guide resource allocation
- Maxwell's equations — near the ground-truth end of the spectrum. physics-grade models that are both useful AND highly accurate.
the spectrum
there's a spectrum from rough useful models to physics-grade ground truth. most real-world models sit closer to the rough end, and that's fine. the mistake is expecting ground truth from a rough model, or refusing to use a rough model because it's not ground truth.
connection to narratives
a narrative is a model of your own story. and like all models, narratives are wrong but some are useful. choosing a better narrative (see confidence) is choosing a more useful model of yourself.
modeling in research
when doing research-workflow, modeling is often the core skill — taking a complex system and finding a representation that's simple enough to work with but accurate enough to be useful.