Generating a novelty scale
post by Three-Monkey Mind · 2019-04-20T09:54:41.400Z · LW · GW · 2 commentsThis is a question post.
Contents
Answers 6 anna_macdonald 3 johnswentworth None 2 comments
I was writing a post when I thought “this isn’t very novel. I should mark it as such.” It then occurred to me to make a 1–10 scale to describe how novel a piece is, similar to Gwern's importance tags. However, I’d prefer an objective scale, not a rank scale that Resorter generates. If I never write anything super-duper-new, I shouldn’t have any 10s on the novelty scale.
Currently, I have a rather coarse four-point scale with points that aren’t uniformly distributed between their neighbors:
- A verbatim copy of something else.
- Demonstrating how to apply a technique publicly documented elsewhere.
- Slightly generalizing an idea publicly documented elsewhere.
- Paradigm-shifting reconceptualization akin to Newton figuring out that the same thing causes apples to fall from trees and planets to stay in their orbits.
I’m trying to flesh out this scale as best as I’m able, but I’d like to get some help if I can. I’m also trying to separate novelty from both utility and importance. For example, “hold the banana by the handle end, pinch the other end to make a crack in it, and peel it starting at the crack” is no more than a 1.1 on the novelty scale, maybe a 4 on the utility scale, and 2 or so on the importance scale.
Are there any other useful points on a novelty scale?
Answers
The problem with a novelty scale is that novelty has a high degree of circumstantial/subjectivity to it. What's new to one person is old hat to another. Millions of people may independently recreate the same wisdom based on their life experiences, and that insight feels new to them, but might not be new to those they share it with. In the modern age, not even a google search can guarantee that an idea hasn't been laid out somewhere by someone.
↑ comment by Three-Monkey Mind · 2019-04-21T00:26:47.588Z · LW(p) · GW(p)
Very true. I think I'm mainly trying to preempt accusations that I'm simply rehashing Taboo Your Words (which I pretty much am rehashing!)
Also, by stating "this isn't very novel", I'm also communicating to the neophyte (as opposed to current rationalists) that there's a wide body of knowledge out there that's quite similar to what I've written. That's potentially useful to the neophyte.
Replies from: anna_macdonald↑ comment by anna_macdonald · 2019-04-22T16:42:40.265Z · LW(p) · GW(p)
Conveying that is often worthwhile, but it's situational enough that simply stating the context of what you're doing is probably a better idea than formalizing a novelty scale.
Also, I didn't mention this above, but re-hashing stuff that isn't novel can be highly useful. Penetration of an idea into the population would never happen if people only ever pointed to the original source for an idea without conveying/spreading it themselves. It's helpful to have a million blog posts about the same thing, because each of those blogs is reaching a slightly different audience.
A general rule that I try to follow is "never write something which someone else has already written better". Rather than give a numerical scale, I'll list a few distinct ways that pieces can satisfy this rule. I'll give examples of each from my own writing, with the caveat that they are not necessarily very good pieces - just examples of my own reasoning regarding their novelty.
- Novel theory/observation/interpretation (or at least novel as far as I know). Ex.: Computational limits of empire
- Carving up known ideas differently, e.g. by pointing to a particular sub-idea or creating a handle. Ex.: The broken chain problem
- Taking a well-known idea from one area, and applying it in another area. Ex.: Competitive markets as distributed backprop [LW · GW]
- Analyzing something which has been analyzed many times before, but usually incorrectly, so it's hard to sift signal from noise. In this case, the novelty/value is in making the analysis visibly correct. Ex.: Accounting for college costs
- Translating some piece of metic knowledge into a more transmissible model/technique. Ex.: How to find unknown unknowns
- Accessible explanations of technical ideas. Ex.: Queuing theory without math
- Arguing that some known idea/problem is more important than people seem to have noticed. Ex.: Coordination economy
Note that these are ways that a piece of writing can be novel, not guarantees that nobody has ever written the same thing before.
Side note: if using a numerical scale, I worry about confusing novelty with importance - the example scale in the OP seems to mix the two. Perhaps a better approach would be to give handles for several different ways things can be novel, and then use those as tags?
↑ comment by Three-Monkey Mind · 2019-04-21T23:36:35.495Z · LW(p) · GW(p)
A general rule that I try to follow is “never write something which someone else has already written better”.
A sensible rule, but I'd like to bring some rationalist insights to other communities that might be able to benefit from seeing how people who've read the Sequences handle things. This seems to necessitate a little bit of redundant writing.
Also, I could stand to get better at writing. On the other hand, if I limit myself to writing only novel things, I wouldn't practice nearly as much as I ought to do. Of course, the decision to publish any given piece is a separate issue.
I worry about confusing novelty with importance—the example scale in the OP seems to mix the two.
Not on purpose. I just couldn't think of something super-novel yet unimportant.
Perhaps a better approach would be to give handles for several different ways things can be novel, and then use those as tags?
That sounds like a good idea inasmuch as it maps to reality the best, but it's also more work than I thought I'd have to do. I'm considering collapsing the novelty scale to no more than five points and trying to make it more coarse to deliberately paper over the different ways a piece can be novel.
Thanks for demonstrating that novelty isn't totally orderable, though; I thought it was, more or less.
2 comments
Comments sorted by top scores.
comment by d0048 · 2019-04-20T15:48:24.064Z · LW(p) · GW(p)
Just had an interesting question: How would you rate the novelty scale of this Novelty Scale concept?
Replies from: Three-Monkey Mind↑ comment by Three-Monkey Mind · 2019-04-20T23:28:27.047Z · LW(p) · GW(p)
On the current four-point scale: 3±ε, where 3 − ε > 2 and 3 + ε ≪ 4. Like I said, these points aren't uniformly distributed.
I'm also familiar with trying to define a unit of enlightenment, so the whole idea of "make a scale" doesn't strike me as a very novel idea.