Your point is well taken, but I need to push back on it. Organizations can in fact understand things in a collective way. Karen Coyle's point about labeling fringe topics is a good example: the organization has to have a system (or "algorithm") for deciding how to present "fringe" content to users. The understanding of what to do can't be delegated to specialists, it needs to be a collective understanding.
"Libraries don't understand algorithms, people do." is a problem statement. Although perhaps my conflation of "systems" and "algorithms" is where we differ. In particular, the algorithms that govern search results and content lables are not expressed in computer code, but in 'people code". > On Nov 15, 2016, at 4:05 PM, William Denton <w...@pobox.com> wrote: > > On 15 November 2016, Eric Hellman wrote: > >> Libraries are participants in this new information enviroment, so I have >> some questions. >> >> 1. Do libraries understand the algorithms and metadata that guide search >> results and suggestions in the services they provide? Do these algorithms >> reproduce biases in our society? > > I try to watch out for conflations of "libraries" with "librarians (and/or > archivists, reference assistants, circulation staff, human resources staff, > maintenance workers, etc.)." Libraries don't understand algorithms, people > do. > > Some of those people are in the administration, some have provosts or mayors > watching them closely; some can speak out without fear of reprisal, others > can't. And when groups get together, a meeting of "libraries" means chief > librarians, but a meeting of "librarians" is very different. > > The rest of your questions are about libraries as institutions, but this > first one caught my eye. Certainly librarians *should* understand the > algorithms and how the search engines work, and the answer to the second > question is, from all I see, yes. > > Bill > -- > William Denton :: Toronto, Canada :: https://www.miskatonic.org/ > Caveat lector.
signature.asc
Description: Message signed with OpenPGP using GPGMail
