@iridescentFluid I find all this very puzzling. When you say free, do you mean as in beer? Or speech?
@mattblaze @iridescentFluid Dude, you don't have to go so hard.
(Also, hi Matt! Good seeing you here. It's been a *while.*)
@mattblaze @iridescentFluid I have been idly wondering whether some of the trust metric stuff I was starting to think about during my internship with you, then developed into Advogato, might help make better moderation tools for something like the Fediverse, but then I think what an incredibly deep pit that is, and go back to making pretty colored pixels dance.
@raph @mattblaze I was thinking that there’s a function of blocking which relates to the diversity of blocks: being blocked by many people with similar lists count less than being blocked by people across a spectrum.
It’s sort of tf/idf?
(Also, Hi Raph!)
@adamshostack @mattblaze Hi Adam, it's becoming a party in here!
@adamshostack @mattblaze I remember the CLEVER Eigenvector stuff out of IBM Almaden. These days, that would give an incredibly strong right/left signal, but I wonder to what extent you could filter that out and get other signal relevant to moderation, such as civility and being willing to listen (as opposed to *splain, for example).
@raph @adamshostack @mattblaze @iridescentFluid DW-NOMINATE is just the first couple of principal components of the voting matrix, right? Maybe you could take the first few principal components of the blocking or following matrix. Maybe you'd get a right-left component but also a "patronizing jerk" component, a "spammer" component, a
"celebrity" component, etc. I think that might operationalize Adam's idea usefully.
@radehi @adamshostack @mattblaze @iridescentFluid This definitely sounds like research worth doing!