Home » Blog » Media » Polymeme and distributed agenda-setting

Polymeme and distributed agenda-setting

A couple of weeks back, I was wondering what functions “we” – which could mean “media geeks”, “people who care about journalism” or “people who believe that informed citizens are important in a democracy” – should try to save from daily newspapers. This doesn’t neccesarily mean saving newspapers from fiscal collapse… though I have to admit, I’m tempted to see if a group of people could raise a million bucks and buy the New Haven Register and hundreds of other smaller papers – the Journal Register Company is down 99% percent in the past two years.

What it may mean is building digital institutions that fulfill some of the journalistic functions newspapers are abandoning. That was one of the motivations behind building Global Voices – Rebecca and I wondered whether citizen media could help fill the gaps in international media coverage resulting from cuts in foreign bureaus and a percieved lack of audience interest in international news.

A great deal of innovation in the media space – led both by new citizen media projects and by the more innovative mainstream media sites – focuses on getting more people to report and opine on the news. There’s been less innovation – or at least, less I can think of – on two other key functions of the newspaper: setting a news agenda, and contextualizing stories for local audiences.

The contextualizing function is a really difficult one to carry out in a digital world. In the newspaper age, smart writers and editors might try to get Boston Globe readers interested in a story about West African cocoa by talking to people in the West African community in Boston, or by talking to high-end Boston chocolatiers. In a digital age, it’s hard to know who’s reading your coverage – or watching your ill-informed rants about China on CNN. Even reading my server logs pretty closely, I’m never entirely sure whether this blog is talking to new media activists in the US, African political junkies on and off the continent, or Italian bloggers with an interest in homophily.

At Global Voices, we try to frame articles broadly, looking for local stories that might be interesting to a broader audience, and attempting to add enough context to make them comprehensible to that audience. Other sites frame by topic – if you’re writing for WorldChanging or TreeHugger, you can assume your audience has interest and basic knowledge about green technology, and that you probably don’t have to explain compost or carbon sequestration. My guess is that the reason that edited aggregators work well (and unedited aggregators tend to be a pretty frustrating experience) is that they’re replacing traditional geographic context with subject and interest context.

(Given the success of expert blogs like Engadget, I wonder how many entrepreneurs are looking at building expert blogs with a tight subject focus, expert editors and highly contextual advertising. I’m fairly desperate for a truly comprehensive African news and opinion blog – someone who’d read the 1000+ top African news feeds and blogs and give a daily digest of the hot stories, like the Morning Brief on Foreign Policy’s Passport blog.)

I’m increasingly of the opinion that the agenda-setting function is the most important and most ignored function we’re losing in moving from bundled, analog media to unbundled, digital media. This isn’t generally a popular stance – there’s a suspicion and hostility to “gatekeepers”, who decide what’s news and what’s not. But gatekeeping is a necessary function – in a world where more than 100 million people are producing online content, it really helps to have someone tell you what you may want to pay attention to, even if it hasn’t already caught your eye. (And it might not be such a bad idea if there was some common set of topics many people agreed were news, if only to create a space for conversation and socialization around those issues.)

There’s no shortage of experimentation in this space, but lots of it focuses on the similar models. Digg and Reddit ask a large group of people to suggest stories and use voting systems to prioritize stories. Systems like StumbleUpon use collaborative filtering techniques that try to match your interests to people with similar interests and suggest sites similar users liked. Both techniques can limit your view of the internet in predictable ways through homophily, the tendency for you to flock with like-minded users and discover content through them. (One of the cleverest ideas I’ve seen on the web recently is LibraryThing’s Unsuggester, which uses the same algorithms and runs them to suggest the content you’re least likely to like… Here’s what Thomas Jefferson probably shouldn’t read. :-)

One of the reasons I’m so excited about Polymeme is that the creators of the tool are trying a different model for setting an agenda. Polymeme looks at a set of “expert” blogs on a particular subject and uses algorithms to determine what stories – in mainstream or citizen media – they’re discussing on a particular day. The Polymeme site features these discussed stories with links to bloggers talking about them, which makes it especially easy to follow conversations that pass between mainstream and citizen media. The stories suggested by the algorithms – but ultimately chosen by human editors – tend to be quite different from those found on Reddit and Digg… or on Google or Yahoo News.

Polymeme is an interesting compromise between a single (usually a team) gatekeeper laying out the front page of the newspaper, and the mob rule – or factional rule – that can dominate a site like Reddit or Digg. It’s explicitly elitist – the stories are selected by watching the tastes of handpicked experts – which is conceptually uncomfortable at first, but I’m increasingly realizing is a very smart strategy. And the stories and topics covered force some interesting reflections on the power and limits of agenda-setting: despite the site’s explicit intention to move beyond overhashed political discussions, Obama and McCain are the two individuals most often mentioned in the blogs the system tracks.

I’d be very interested in hearing other examples of projects that are building novel systems – automated, human or a hybrid of both – to set news agendas. Are there models beyond the three listed here: trusting an expert, trusting your (realworld or collaborative filtered) friends, trusting a mob? Who’s implementing those methods well?