To understand, study edge cases
post by Shmi (shminux) · 2019-03-02T21:18:41.198Z · LW · GW · 8 commentsContents
8 comments
This seems obvious, but is often forgotten. When a system functions efficiently, its individual parts are hard to notice. If you want to understand how a system functions, what it is made of, how each part works, but you cannot easily disassemble it and have a look, you have to get out of the nominal regime and let each part function independently, as much as possible.
Some examples:
- To understand subatomic particles in a nucleus, or to figure out the structure of atom, hit them hard with other particles and see what happens.
- To understand free will, focus on the situations where the notion fails.
- To understand consciousness, study altered states.
- To understand mental health, study mental illness.
- To understand anatomy, dissect cadavers.
- To understand logical reasoning, study biases.
- To understand evolution, study artificial selection.
- To understand economics, study market failures.
8 comments
Comments sorted by top scores.
comment by quanticle · 2019-03-03T00:06:18.010Z · LW(p) · GW(p)
There are fields where studying edge cases leads to confusion and actually hinders progress. From gwern's excellent essay on Bakewell and the origins of genetics:
But surviving theoretical scientific discussions of heredity are baffling. People lurch between ‘only fathers matter’ & ‘only mothers matter’, endlessly elaborating on wildly speculative (and wildly wrong) mechanistic explanations of how exactly sperm & eggs & embryos connected and formed, and in an example of “hard cases make bad law”, the focus on ‘monsters’ and other extreme cases among humans or animals badly misguided their premature attempts to elucidate universal principles comparable to that of astronomy or physics
The lesson is that when attempting to study statistical effects that aggregate across populations (like with genetics), studying the edge cases will lead one away from truth rather than towards it. Bakewell, Mendel and Darwin didn't develop their theories of heredity and genetics by studying plants and animals deformed by mutation. They studied populations of "normal" plants and animals, and kept very careful records of the statistical rate at which characteristics were transmitted from parent generations to child generations.
Replies from: shminux↑ comment by Shmi (shminux) · 2019-03-03T05:12:31.234Z · LW(p) · GW(p)
I agree that edge cases are not very good for synthesis. They are essential for analysis, however. Noticing patterns in a well functioning system is useful for building phenomenological models of the system, but not necessarily for figuring out its constituents. And yeah, sometimes it is possible to figure out a lot from a nominally functioning system.
comment by Aiyen · 2019-03-03T17:43:41.685Z · LW(p) · GW(p)
"To understand anatomy, dissect cadavers." That's less a deliberate study of an edge case, and more due to the fact that we can't ethically dissect living people!
Replies from: shminux↑ comment by Shmi (shminux) · 2019-03-03T19:06:18.548Z · LW(p) · GW(p)
I meant it more in a sense that people like Michelangelo had to understand the inner workings of the human body, well, the anatomy of it, to paint it faithfully when it's in one piece.
comment by Charlie Steiner · 2019-03-03T14:00:29.777Z · LW(p) · GW(p)
Is this contra https://www.lesswrong.com/posts/aNaP8eCiKW7wZxpFt/philosophy-as-low-energy-approximation [LW · GW] ?
To repeat my example from there, to understand superconductivity it doesn't help much to smash them into their components, even though it helps a lot for understanding atoms. A non-philosophical example from your list where people went to the "extremist" view for a little too long might be mental health before the rise of positive psychology.
Replies from: shminux↑ comment by Shmi (shminux) · 2019-03-04T08:14:27.669Z · LW(p) · GW(p)
Not sure if it's "contra" your post. In physics there are effective theories and the Intentional Stance is an example of one, which is, indeed, a "low-energy" approximation. My point is not about energy level, but more about finding the limits of applicability to see the constituents better. For example in a control system it is hard to figure out the innards until you, say, open the feedback loop in some way, or find a regime where the the system goes into resonance, or saturates, etc.
comment by Matt Goldenberg (mr-hire) · 2019-03-04T17:19:15.016Z · LW(p) · GW(p)
Good concept.
There's a corresponding principle in design, which is to design for extremes. If you want to make a pair of scissors easy to use for the average person, think about how to make them easy to use for both a person without a thumb, and a person with only a thumb.
comment by ChristianKl · 2019-03-05T15:04:58.115Z · LW(p) · GW(p)
To understand anatomy, dissect cadavers.
A lot of our view of anatomy comes from the knife. As a result many people consider assumptions that are true for cadavers also to be true for living tissue.
Living bones with fascia on them happens to be a lot more dynamic then commonly assumed.