Abstractions allow us to quickly and easily think about things that might otherwise be laborious or even possible.
In this way, abstractions can make us smarter.
Unfortunately, when we form an abstraction, we abstract away some of the details, and occasionally those details matter.
One type of abstraction is the formation of a category.
Last week, two categories made bright people in a bright organization stupid in a way they otherwise would not be.
I have somewhat elevated triglycerides but low cholesterol. This makes statins incredibly unlikely to be useful to me. I have genetics that put me at high risk of having adverse reactions to Stains.
The doctor has a single category of “people with high triglycerides,” which is typified by people with triglycerides that are elevated 3-fold as much as mine, have similarly elevated cholesterol, and no genetic issues with statins. So he enthusiastically suggested that I start on statins because they are (and I quote), “the frontline drug” to fight triglycerides. There are four other reasons why this drug would be a bad idea for me (two of which apply to most of his patients with elevated triglycerides), but I am in the category.
The doctor also suggested that I cut “processed meats” out of my diet as a key part of triglyceride control. He has a category of “processed meats,” which is bad for triglycerides. He isn’t even aware he that he has a category. While this, in fact, is true of most processed meats, it is not true of the ones that I (and most of his patient group) actually eat. Because of this categorization, he is suggesting a difficult unnecessary lifestyle change.
If I had been eating the “bad types” of processed meats, it would be simple to switch to healthier ones. It wouldn’t have been a lifestyle change that we couldn’t have easily changed, it would be a note on my shopping list.
I have concierge-level doctors with plenty of time to spend with me, access to the best test equipment, etc., but since categories had invaded their minds, the gave me (and presumably countless others) bad medical advice.
Sometimes abstractions are powerful tools for thought, but they always carry the risk of making us stupid in new and interesting ways. It’s an intrinsic risk.
When we think about machine-leaning systems, stereotyping, and societal prejudice, we are similar categorization problems, but in situations that are more emotionally fraught.
Those are much harder problems. Perhaps we can start to understand those systems with this much simpler problem where everyone is of goodwill, acting in good faith, on the same, working toward a shared goal.