Error

So let me know in the comments if you’re interested in a followup post on *how* to build models.

Yes please!

Great to have this written up. I usually had this experience when talking to senior professors; they tend to have so much cross-referential knowledge that any piece of data just does much more work than for non-experts. I just never explicitly looked at expertise from a model-perspective. Your posts gives good pointers, but are there any things you see good model-builders intuitively do better? Like not just reading but questioning and grokking. I feel like model-building is also not fully learnable but somewhat distributed among people so I’d be interested in possible differences.

Endnote 2: Why can’t we just defer to existing experts, instead of figuring stuff out for ourselves?

Alternative & complementary response: which experts? Why them, instead of these other experts who disagree with the former? How can you tell if you’re (say) being misled? To quote John Wentworth:

When non-experts cannot distinguish true expertise from noise, money cannot buy expertise. Knowledge cannot be outsourced; we must understand things ourselves. …

King Louis XV of France was one of the richest and most powerful people in the world. He died of smallpox in 1774, the same year that a dairy farmer successfully immunized his wife and children with cowpox. All that money and power could not buy the knowledge of a dairy farmer—the knowledge that cowpox could safely immunize against smallpox. There were thousands of humoral experts, faith healers, eastern spiritualists, and so forth who would claim to offer some protection against smallpox, and King Louis XV could not distinguish the real solution.

John also suggests that the kind of deep model you want to build is gears-level models (that link has a lot of examples across various domains):

If I want to build long-term knowledge-wealth, then the analogy between money-wealth and knowledge-wealth suggests an interesting question: what does a knowledge “investment” look like? What is a capital asset of knowledge, an investment which pays dividends in more knowledge?

Enter gears-level models.

Mapping out the internal workings of a system takes a lot of up-front work. It’s much easier to try random molecules and see if they cure cancer, than to map out all the internal signals and cells and interactions which cause cancer. But the latter is a capital investment: once we’ve nailed down one gear in the model, one signal or one mutation or one cell-state, that informs all of our future tests and model-building. If we find that Y mediates the effect of X on Z, then our future studies of the Y-Z interaction can safely ignore X. On the other hand, if we test a random molecule and find that it doesn’t cure cancer, then that tells us little-to-nothing; that knowledge does not yield dividends.

John has some advice on how to read papers to build gears-level models, although for most situations I prefer Sarah Constantin’s advice to do fact-posting.

What are the sources for your examples? Particularly interested in Al Smith but I imagine other people might be interested in the others as well.