|
Curation for discovery has been the most difficult problem for information since access exploded with the internet. Netflix had prize money behind algorithmic improvement here. Public libraries and physical bookstores use experts and physical placement. Most digital content services use humans to curate new content. The last chapter in James Gleick's book "The Information" identifies this as the next problem in information - now that access is no longer a barrier, how do we decide what's useful? The combination of wikipedia and Google has largely solved search for most internet-available content, but not discovery or trust of what you don't yet know to search for. This curation problem isn't only about blog content - it also contributes to the proliferation of fake news and conspiracy theories. Comment curated by many people's networks has become limited to their tribal connections, and some tribes exclude information sources that value scientific or journalistic standards. What we already know about and read daily is what we trust, and helps define our tribe. Our tribe uses HN. Other tribes (which may intersect with ours) use Reddit, Digg, 4Chan, forums, newspapers, blogrolls, and many other systems. When we get content from our tribe - the concept of 'social media' - someone else is sharing their trust in content with us, in ways we can categorize. Each content system on the internet is rooted in a point source, the owner, who shares their trust with varying levels of directness: Newspapers don't have a pattern of trust sharing, though some experiment. Blogrolls' owners share trust directly - 'direct owner trust sharing'. Small forums (where content is only ranked by time, not votes) use 'direct user trust sharing' because users are treated equally; Reddit posts use 'indirect user trust sharing' through voting. Facebook and Twitter use both 'direct user trust sharing' (what your friends post) and 'indirect owner trust sharing', (the owner algorithms). It seems clear that there are serious manipulation problems with indirect trust sharing systems open across tribes - Reddit, Facebook, and Twitter all struggle with this. Innovation in social media seems to focus on curating direct user trust sharing to guide indirect user trust sharing. To fix trust sharing, we'll have to curate new direct relationships. I believe this will be much along the lines of blogrolls - direct trust sharing from identified tribal members, rather than only from your direct network. Work has been done to map the internet through linking. Some of this, such as Jonathan Albright's coverage of the 2016 US Presidential election, has begun to identify tribes in those maps. I believe the next major algorithmic innovation will triangulate your tribe, using content you've already trusted yourself (by sharing with others), and curate discovery of new tribal members based on the nodes in and near that tribe. A network-wide blogroll, across content types, which links you to new direct trust sharing relationships. This would help democratize curation, and could result in a clearer understanding of tribal boundaries and how disinformation spreads. |
Back in 1989, when Byte Magazine was warning us of "infoglut", I decided that the most valuable job was the editor. I've seen no reason to change my mind since.
I ran the hub of a BBS network, moderated numerous forums, have run a few user / study groups, published a few zines / newsletters, was a (terrible) radio DJ, etc. I even work on recommenders for my day job.
I remain unimpressed. I've tried to be open minded about automated curation. But I've decided recommenders, as we know them, have topped out.