FriendFeed co-founder Bret Taylor has a nice interview with CNET’s Dan Farber today. In the interview, Bret mentions that FriendFeed will be introducing ranking algorithms soon. These will create your own personalized FeedMeme. Which is going to be interesting. Dan Farber then asks whether user controls over content will be rolled soon. Bret answers ‘no’.
In the discussion around this link to the interview, Kingsley Joseph comments:
great interview. Bret’s got it right – ranking algorithms, not filtering is the key to noise processing
This was an interesting comment. I think of noise as a very personal thing. And the ability to define your own take for what constitutes noise makes sense to me. But Kingsley has a different point of view.
So I wondered…are we talking about the same thing. What exactly is noise?
A Definition of Noise
The diagram below is my definition of noise.
Signal is all about the stuff you actually want. For some, it’s a steady stream of social media entries. For others, it may be a steady stream of parenting items. Or baseball discussions.
Discovery is the middle ground. It’s things you weren’t looking for, but find interesting. Maybe Flickr Favorited pictures. An interesting tidbit from the world of science. An Inquisitr celebrity update.
Noise is the stuff you just don’t have the patience to put up with. It’s not anything you’re seeking, it’s annoying you that it’s even on your screen.
In this definition, there’s a fine line between discovery and noise. I’m argue that its the quantity of entries that determine whether something is noise or not. A few items creeping into your FriendFeed is probably all right for all but the hard core signalists. At some point though, when the volume of stuff you’re not seeking crosses a threshold, the entries become noise. You’re not getting enough of the stuff you’re seeking.
Everyone has their own threshold for discovery versus noise. This is the personalization required for noise control. Alexander van Elsas touches on this issue in a recent post:
It’s the noise problem (Try a search on “noise” here for example). How can we find the things that are really important from that huge pile of information floating around. That is partially why we have aggregation and filtering services. Each of them, using one algorithm or another, tries to compile a tiny subset of the universe and present that to its users. The question that remains is whether or not the right tiny space is presented.
Alexander strikes me as being closer to the signalist end of the information spectrum.
FriendFeed Rooms + Ranking Algorithm > Filters?
This brings me back to Kingsley’s point of view that ranking algorithms are optimal for noise control. Ranking algorithms are absolutely terrific. I love them. They provide a lot of benefits in a number of areas (e.g. Google Search).
I’m should also note FriendFeed’s Rooms as the other initiative for isolating topics and controlling noise.
Are they enough to control the noise? It’s hard to say yet. For Rooms to be effective, members are going to need to use them pretty frequently. That can happen over time, but it’s still dependent on the broadcasting member using them. There won’t be 100% compliance because:
- I can’t perfectly predict what my subscribers will think is noise
- Likes and Comments on something outside my usual programming put entries into my subscribers’ streams
The ranking algorithms will be interesting. Looking forward to seeing what gets rolled out. Knowing the FriendFeed gang, it’ll be good.
The difficult part is finding the middle ground between the hard core signalists and the discoverers, who like a fair amount of entries outside the stuff they seek. Neither group wants noise, but the threshold for the number of unseeked items varies greatly. My tolerance is pretty high, so I guess you’d call me a discoverer.
Balancing the desires of the signalists and the discoverers, and the varying thresholds of individuals will be the challenge. Let’s see what the ranking algorithms do.
I’m @bhc3 on Twitter.