Hey #CompSocSci and #DigitalHumanities people, do any of you have experience with studying Tinder? Is it even possible to collect Tinder data? Any hints (and boosts) appreciated!

Out of curiosity, for (a) scientific research, (b) potentially investigative & (c) historical purposes, how is archival for the entire (growing) fediverse, not just mastodon, being/going to be handled? I’m guessing this is different from platforms like reddit or twitter since those are more centralized. Should it be done? Would it be handled per instance? Should there be opt-in or opt-out for each instance/user? #internetarchive #fediverse #fediquestions #fedihelp #socialscience #compsocsci

potentially naive #question for the #NLProc and #CompSocSci crowd.

I'm doing gradient boosted tree-based regression with social media engagement metrics (e.g., # retweets) as outcome variable. Which means overdispersion + lots of mass in the low end of the distribution.

Seems like Tweedie loss may be the silver bullet for this kind of outcomes. Any downsides or alternative approaches I should be aware of?