| > - How to be healthy - How to educate - How to raise kids I think it's interesting that all of three of these hinge on the same problem: human variability. I strongly suspect that humans are much more variable internally than conventional science and journalism accounts for. The scientific method uses statistics which is important for showing aggregate effects but can also obscure variability. And journalists have an incentive to offer simplified guidance that is as widely applicable as possible. The end result of a lot of information like "Do X" when what it should really be is "If you are like Y do X, if you are like W do Z." This is a pattern I think of as "the missing parameter" that I see just about everywhere people give advice once I started looking for it. Examples: - "(Static/dynamic) types don't help programming." For what kinds of programs? Developed by whom? At what scale and timeframe? - "You (should/should not) use a schemaless database." How big is your dataset? What is the relative fraction of reads/writes? What kind of data? How many users? How much money do you have? What are your failure modes? - "People should move (to/out) of cities." Which cities? What is the city's transit like? Which people? What activities do they prefer? Do they have kids? What's their economic status? Age? How important is it to be near a hospital? - "You should eat less salt." For people of what age? Activity level? Diet? - "Teach by example not generalities." Teach what material? To whom? How important is their understanding? How long should they retain it? - "Children should go outdoors on their own more." What age? What personality do they have? Where do they live? Whenever I see people arguing about some generality, what is often happening is that each party has implicitly filled in those parameters with different values, so they are all correct but talking past each other. When I see this happening now, I try to take a step back and figure out what implicit parameters they are assuming and see if there's a higher-level parameterized stance that unifies their arguments. |