So the question is “How actually do scientists gain knowledge, once we admit that there is no concise a priori answer? And how can we find that out?”
-
-
Replying to @Meaningness @s_r_constantin
One obvious approach is to ask them “how did you determine this specific fact yesterday,” and then they launch into a story about chromatography columns and ethidium bromide or whatever.
1 reply 0 retweets 3 likes -
Replying to @Meaningness @s_r_constantin
Then instead of trying to turn that story into a tidy morality fable about The Scientific Method, you can take it seriously in its own terms. What specifically *is* the logic whereby that experiment shows protein A regulates protein B.
1 reply 0 retweets 1 like -
Replying to @Meaningness @s_r_constantin
Another thing you can do is to hang out in labs watching scientists do science. Then what you see is “shop work” that is almost perfectly dissimilar to the fables you are taught in HS/undergrad about how science is done.
1 reply 0 retweets 1 like -
Replying to @Meaningness @s_r_constantin
The actual work is mostly improvisational futzing around with materials and equipment, trying different things out, trying to coax them to produce an answer. And when you do that, you run into the “contingencies” Garfinkel enumerates.
2 replies 0 retweets 2 likes -
Replying to @Meaningness @s_r_constantin
Phil’s insight was that the contingencies are constraints on the form of a cognitive architecture.
1 reply 0 retweets 1 like -
Replying to @Meaningness @s_r_constantin
E.g. if you assume knowledge consists of datastructures representing fopc wffs, you inevitably hit a combinatorial explosion. So we applied modus tolens, and concluded that knowledge can’t be datastructures or wffs or anything like that. Our program Pengi did fine without them.
2 replies 0 retweets 0 likes -
Replying to @Meaningness
ah, the thing where you can't proceduralize a scientist. (or an engineer or mechanic for that matter.) yes, that's quite true.
2 replies 0 retweets 1 like -
Replying to @s_r_constantin @Meaningness
"shit, the mysterious phenomenon was due to a difference in the manufacturing process" comes from outside, it's not like you had a preexisting model with a node for the manufacturing process. you don't have a sample space, you have to have "room" to "add stuff from the void".
2 replies 0 retweets 0 likes -
Replying to @s_r_constantin @Meaningness
That's why a single probability model doesn't cover everything; you keep having to respond to new stuff from the Void and change the ontology of "what do I mean by an event in my sample space."
1 reply 0 retweets 1 like
Yes! Yes yes yes! This is what “meta-rational statistics” means, and is a major topic of my book!
-
-
Replying to @Meaningness
(I'm not sure that there aren't ways to translate this into a Bayesian universal prior, but at any rate, if I were trying to program this legibly, I'd have a finite model plus an object I called "Void" that sometimes threw out random shit, not a universal prior.)
0 replies 0 retweets 0 likesThanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.