Nicholas Carr over at the Rough Type blog (really worthwhile blog, btw) argued yesterday that our information overload problem isn’t a filter failure problem (i.e., too much noise and not enough signal) but a signal problem (i.e., our filters are so good they are overloading us with signal).
Carr’s argument, in a nutshell: our filtering tools are good enough that finding a needle in a haystack, where there’s a body of information and we need something specific (“situational information overload”) isn’t a problem. The problem, he argues, is experiencing overload because we are surrounded by too much information we are immediately interested in (“ambient overload”), and that this is the result of the filters being so effective at eliminating noise that they overload us with signal.
I think he got some of it wrong and some of it right. His distinction between situational and ambient overload is pretty interesting and useful, and I think is analysis of situational overload is all right, but I disagree with his assessment of our filtering tools. I don’t think they are all that sophisticated, in fact. Our filtering tools work fine for simple searches, which is what he seems to be talking about. You can pretty easily find the relevant court cases on an issue or point of law, for example, find relevant web pages through Google or Bing for a specific set of search terms, watch a conversation on Twitter around specific hashtags, or track a specific group of people’s Facebook posts.
But complex filtering is another matter altogether. For instance, if you are working on community decision-making process design and want to use search and filter tools to track information and conversations from other design fields that bear on the challenge, good luck to you. Within narrow, tightly defined universes, fine, but start crossing universes and the tools quickly get sloppy.
And I think he’s totally wrong about ambient overload. It’s not that I think there isn’t a too-much-signal problem – I like that part of his analysis – but there is also a too-much-noise problem. Ambient is almost the same as noise by definition, with plenty of signal but even more noise. Most of what passes on my Facebook stream – ambient by his definition – is noise, with bits of signal mixed in. Don’t care about that, don’t care about that, don’t care about that, not that either, oh wait that’s cool. But it’s seductive, that stream, and I want to keep watching, waiting for a hit of something worthwhile. For Facebook to be really useful to me, I do need better filters that can figure out what sorts of info I’m actually interested in and block out the rest.
Facebook actually just took a stab at this but with characteristically poor execution. Their approach (only show you wall posts by people you interact with on Facebook) created a self-reinforcing feedback loop that doesn’t distinguish on the basis of information quality at all, and in fact makes it really difficult to ever encounter the serendipitous surprises that make Facebook delightful at times. (You can turn this feature off, thankfully). Ditto on Twitter, RSS, actual news feeds, my email inbox, and so on. Maybe the signal:noise ratio is better with ambient, but it’s a mistake to presume that means the problem is one of too much signal.
I think one problem may be that his notion of filters is too simplistic, and those sorts of filters probably can’t solve the overload problem very well. The very rich idea of curation is pretty helpful here . . . it’s not so much filtering that we need but information curation to help us see the stuff we most need to see and help us understand its relationship to other information. Better automated curation tools might help a lot, but in the meantime I don’t think there’s any way around the critical role of human curation. I don’t mind that at all, and finding and sharing the most important conversations and innovations on decision design and support is a growing part of what I do at PlaceMatters. It’s also part of why I so enjoy finding others who are particularly good at information curation, since that’s a critical tool for making sure we learn from ideas across sectors. But it’s solving a noise problem, not a signal problem.
Finally, his analysis entirely overlooks what I think is a particularly critical element in intelligently and joyfully managing information and content: mechanisms that facilitate serendipity, unexpected but high-value information you would not have found otherwise. Curation can be a powerful strategy for embedding this, just as good design can.
Overload is obviously a problem, and incremental improvements in filtering tools probably can’t keep up with the exponential increase in information flows. But I suspect the answer really does lie in the evolution of filtering as a concept (and then the tools that follow) rather than in “prayi[ng] for filter failure.”