On a busy day, Andy Carvin gets 2,000 Twitter @-replies.*
Which, wow. We all experience, in some way or another, information overload; Carvin experiences it on a whole different level. And while he’s able to do the work he does largely because of the institutional leeway he’s given to do it, the work of curating the social web also suffers from a fundamental problem: @acarvin, the account, is run by Andy Carvin, the guy. The guy who, as smart and quick and well-sourced as he is, has a 24-hour-long day just like the rest of us.
Though Andy Carvin can’t scale, however, @acarvin might be able to. Which is an idea that Carvin himself, along with Jeff Jarvis, explored in an unconference session that just wrapped up at today’s MIT/Knight Civic Media Conference. A session titled, only slightly cheekily: “Leveraging @acarvin.”
In their discussion, Carvin and Jarvis identified two broad — and potentially contradictory — needs for the work Carvin does: efficiency and veracity. On the one hand, given the flood of incoming tweets he deals with, and the crucial information that some if not many of them contain, Carvin needs to find a way, he noted, to up his feeds’ signal-to-noise ratio — and to do it in a way that won’t find him constantly drowning in a sea of @-replies. And on the other side of things, he needs a way of tracing both truth and rumors as they travel along on the currents of the social web.
And NPR is thinking of ways to help him do both. Carvin, during today’s talk, mentioned some intriguing tools that he and his NPR colleagues are thinking of building in the service of “leveraging @acarvin.” Two of them:
Triaging @-replies. Given the thousands of reach-outs that come to Carvin’s feed every day, he needs a way of sifting through them to find the good intelligence. The solution could be the creation of an algorithm that would essentially cross-reference the people who reach out to Carvin on Twitter with the social graph of his existing sources.
That process would result in something like a Klout-like score for sources, Carvin said, but tailored to the real-time nature of his work. (Klout itself isn’t ideal for the kind of work he does, Carvin noted — which might find a protest participant with, say, 15 followers offering him the most valuable, and accurate, information about a given event.) What Carvin needs is a kind of transitory Klout for potential sources: “Do they have influence with the people I’m interested in at this particular moment? Have they geolocated their tweets? Do their tweets have links to a YouTube video or a Flickr photo or a Twitpic?” All of those could potentially be factors in determining @-replies’ immediate relevance to him, Carvin said.
“And the output for this, I think, would be a series of Twitter accounts,” Carvin noted — which would mitigate the need for, say, designing a new Twitter client. “And so, if [the algorithm] is churning out all these people it thinks I should be paying attention to in the @-replies, I could just subscribe to that as a Twitter list or a Twitter account” — and then access that source list across platforms.
That algorithmic assist wouldn’t mitigate the need for the kind of hands-on sourcing Carvin does in his work; but it would add much-needed efficiency to that work’s process.
Meme-tracking tools that are tailored for journalism. There are some that are already out there, Carvin said, “but none that I’m aware of that are specifically targeted toward journalists.” During the Mumbai terror attacks of 2008, Carvin noted, a rumor spread on Twitter that terrorists were using social media to coordinate their actions — and there was a rumor on top of that that the Indian government had told its citizens to refrain Twitter. “So there was a period of several hours,” Carvin noted, “where some of the best eyewitnesses got skittish.” From what we can tell, he said, the rumor likely started on TV, and found its way to Twitter. “But in situations like that,” Carvin said, “I’d want to be able to know, in real time, ‘Who started this?’”
Same deal for, in Libya, the recurrent rumors that Muammar Quaddafi is dead. “Is it wishful thinking on someone’s part?” Carvin asked. “Is there a group of people, or a single person, who’s constantly putting this out there?” While questions like that aren’t always hard to discern — most of the disinformation that Libya and other states put out “is so flagrantly obvious” in its falseness — “at some point, they’re going to get more sophisticated.” It would be great to have a way of tracking the proliferators of rumors, he said.
So those are two approaches, notional but exciting. For all the enthusiasm surrounding the idea of scaling the work Carvin does, the question of how to do it tends to get stuck on a persistent problem: the fact that @acarvin’s power as an agent of newsgathering is based on Andy Carvin’s power as an agent of newsgathering. Carvin’s rich knowledge of the Middle East — the sources he’s been able to cultivate based on a pre-existing network of friends and acquaintances in the region — aren’t things that can be easily taught, or transferred, or otherwise replicated. But it’s good to know that NPR, like many media watchers, is considering questions of scale when it comes to Carvin’s work.
“None of these are emergency hacks at the moment,” Carvin noted. Rather, “these are things that would help over the long run — and that I think would help others.”
*This graph initially said that Carvin’s @-reply rate on an average day was 2,000 — and that, after getting a big media mention, that number would jump to 10,000. I misunderstood him, though; he’d meant that, with the increased media attention on it, his account (which currently gets 2,000 replies on a busy day) has the potential to garner 10,000 @-replies — which is why he’s thinking about scaling it in the first place. Sorry about that; I’ve removed the mention of the 10,000 replies in the post.
Image by personaldemocracy used under a Creative Commons license.