"formal" here, hmm think MIRI style try to prove mathematical theorems instead of just philosophical arguments.
Conversation
this is not to denigrate philosophy btw, I think LW ppl. often do premature formalisation.
2
3
Have you met Joel?
Quote Tweet
Replying to @JoelChan86 and @chinmay
I think incremental formalism is one answer.
1
4
Incremental formalism is best term of art description I've heard in #roamcult since 's accretive knowledge work + programmable attention
brought the theory into the community
Quote Tweet
6/ @JoelChan86's article on synthesis and "incremental formalization" at @roambrain has been very helpful in clarifying what I've already doing for years, just in the arts instead of science, and which cruder tools until finding @RoamResearch a while ago.
roambrain.com/knowledge-synt
Show this thread
3
4
21
Cool! I've seen his zettelkasten interview, but haven't read any his other research on knowledge work etc.
1
Latest work on synthesis is still brewing. One decent entry point is here: terpconnect.umd.edu/~xinq/CSCW19_k (predates my encounter with Roam, but the overlap in vision is remarkable).
Also check out if you haven't yet, lots of great entry points into ethnomethodogy.
2
9
You'd shared this paper before but somehow I'd totally missed the intersection between "grounded claims" and note-titles-as-API. I actually paid attention this time and really enjoyed it! Thanks, Joel.
(my extracted notes…
notes.andymatuschak.org/Grounded_claim )
3
5
28
To your points, I agree there is significant overlap in those concepts. We have been softening our ideas about grounded claims to include other more granular units like concepts. But the key idea of needing it to be contextualized remains.
2
1
9
One thing I was curious about was: why the 1:1 relationship between claims:papers, rather than many:many? If a claim is made in many papers, I make many distinct grounded claims and link them, yes? Seems like a clear decision point, so I bet there’s a good reason.
1
5
Huh, I missed that: we didn't mean to imply 1:1 - many:many makes way more sense (exactly as you say), and is actually what we do. Wondering now what we wrote that gave that impression? Want to be clearer in future explanations.
1
The diagrams and workflow description in the text are all 1:1. Actually, I’m not quite sure how (given the description of the interaction) one would associate a summary with multiple distinct PDF excerpts!
Ah you're right! Not described in paper, in part bc it wasn't polished: can create excerpts by themselves, and link multiple excerpts to a single claim. Can also link existing excerpts to other claims. I've done this, eg, to associate a graph + methods excerpts to a claim.
1
1
Ah, got it! Great! Thanks.
1
Show replies