Conversation

Notices

1. Embed this notice
   Sick Sun (sun@shitposter.world)'s status on Thursday, 21-Nov-2024 05:23:55 JST Sick Sun
   I don't really know what I am talking about but I swear I read a paper a couple years ago that was about detecting bias in data, I mean the theory that all data is biased kind of makes sense but natural distributions unreasonably conforming to bell curves seems to be a real thing
   • Haelwenn /элвэн/ :triskell: likes this.
   • Embed this notice
     Sick Sun (sun@shitposter.world)'s status on Thursday, 21-Nov-2024 05:26:08 JST Sick Sun

     in reply to

     • tiskaan
     @tiskaan that's a tautology
   • Embed this notice
     tiskaan (tiskaan@netzsphaere.xyz)'s status on Thursday, 21-Nov-2024 05:26:09 JST tiskaan

     in reply to
     @sun bell curves are natural distrobutions
     BowserNoodle ☦️ likes this.
   • Embed this notice
     Sick Sun (sun@shitposter.world)'s status on Thursday, 21-Nov-2024 05:31:07 JST Sick Sun

     in reply to

     • Niggy Hardare
     @J the issue I have with the argument is that it often is a sleight of hand from "everything is biased" to "all bias is equal" like if you make any measurement, in the universe there is some other completely contradictory measurement that disproves some observation. Basically if I am trying to learn from data and someone's fallback is an epistemology argument that you can't know anything, I just want to punch them as hard as I Can
   • Embed this notice
     Niggy Hardare (j@cawfee.club)'s status on Thursday, 21-Nov-2024 05:31:08 JST Niggy Hardare

     in reply to

     @sun Not surprising, that problem has more or less been predicted since Descartes. There's no way to decouple a raw fact from the interpretation of that fact.

     We have to transcode all external information into a format we can comprehend, and we change it in the very process of doing that, and the process we do that by is so complex that everyone is creating a new thing each time a point of data is observed, and that might be fundamental to how the world actually works at the quantum level.

     It kind of fucks up everything we naturally try to do.

   • Embed this notice
     Sick Sun (sun@shitposter.world)'s status on Thursday, 21-Nov-2024 05:40:08 JST Sick Sun

     in reply to

     • Niggy Hardare
     @J while I took extensive philosophy courses in my undergrad but I am still basically an engineer brain. the tons of cases you find an error or limitation in observation and you update your collection process and you don't have to re-evaluate your philosophical model of the world. yeah it is critical you need to be able to update your model but that's not the first thing you modify. in fact in some cases people want you to change your model because they are upset it's too good at predicting things and they don't want to update their model
   • Embed this notice
     Niggy Hardare (j@cawfee.club)'s status on Thursday, 21-Nov-2024 05:40:09 JST Niggy Hardare

     in reply to
     @sun It's very annoying but it's also completely true that you (or any particular person) are more or less just inventing a tentative reality that might not reflect anything true, and the fact that you can't remove bias from interpreting data reflects that. It's useful for knowing how much to trust yourself, and how seriously to take others, because they might literally be living in a different universe from you.
   • Embed this notice
     Niggy Hardare (j@cawfee.club)'s status on Thursday, 21-Nov-2024 05:44:43 JST Niggy Hardare

     in reply to
     @sun Only because we have to be limited by our attention. We don't have the capacity to create new models every time, it's "good enuf" until it isn't. We have to be efficient.
     
     All that just to say, you will never get pure, unbiased data in any category, ever.
   • Embed this notice
     Sick Sun (sun@shitposter.world)'s status on Thursday, 21-Nov-2024 05:44:43 JST Sick Sun

     in reply to

     • Niggy Hardare
     @J I think that in any circumstance the heuristic is to change as little as possible that gets you back on the right track if you detect a problem.
   • Embed this notice
     tiskaan (tiskaan@netzsphaere.xyz)'s status on Thursday, 21-Nov-2024 05:49:05 JST tiskaan

     in reply to

     • Niggy Hardare
     @sun @J yeah you just test the data against your models and even ask yourself if your biased too
     Sick Sun likes this.