Notableness

This was originally in response to an email from Eric about the noise and relative value of Flickr photos in his friends’ photo streams.

I like the concept of notableness. Certain criteria of the datum fall outside the bell curve, warranting more attention.

We dealt with this issue on Vox—a linear stream of equally weighted items. In reality, items are not equal. Not only that, but different items are more notable to different viewers.

This morning I unfollowed someone on Twitter because she spammed the front page with several tweets per hour. This is akin to the nuclear option, something software should solve, really. Typically, her tweets are low value to me, but in aggregate they are interesting. Or if she has one that’s well-replied to or starred, it is isomorphically more significant.

Lastly, an algorithmically condensed event stream isn’t something that nicely fits into an RSS feed. Or maybe. This is where the rubber meets the road. Capturing—and therefore limiting—a set of events into an aggregate event, then freezing it in time as a discrete RSS entry might be a reasonable compromise between theory and reality of implementation.

Comments

October 1
11:56 AM
Edit

kellan writes:

Freezing this stuff into a feed item is going to be largely impossible from both a UI standpoint but also just for the cycles burned to build it. I think this is where we start to see the break down of the stream interfaces -- feeds fading back into obscurity as niche aggregators fish from them.

October 1
12:21 PM
Edit

ydnar writes:

I think a feed representation of this data is necessary, if only because of the existing feed reader inertia, and because Yet Another Inbox doesn’t scale.

If you break the presumption that feed items, once spat out, are immutable, then this starts looking more realistic. This would require some retooling on the part of the feed readers, and/or extensions to feed format(s), but neither of those are show stoppers.

Reusing the Flickr example—because notability maps pretty well to interestingness—a user uploads 100 photos, and this is condensed into a single event item. Then, one photo gets lots of views/comments/favorites, and when the user’s feed is evaluated again for aggregation, this causes a distinct new event item to spring into existence.

Where this gets interesting is evaluating the user’s feed in multiple contexts—standing alone, or in the context of a follower/friend. For example: mutual friends viewing/favoriting as a factor in determining notability for a given discrete event.