John Kelly, Morningside Analytics on the fact checking ecosystem

John Kelly, chief scientist of Morningside Analytics, makes pretty diagrams that feature multicolored dots. The pretty dots frequently tell complicated and subtle stories about the spread of ideas in online media spheres, particularly the blogosphere. (Tragically, I don’t have Kelly’s slides for the talk, which means I’ll be trying to channel a very visual talk here…) He maps social media citations and studies the resulting topologies to understand the spread of ideas.

To understand what conversations are taking place about fact checking, he takes a “semantic slice” of the network. He looks for markers – keywords, URLs and metadata – and offers a “relevance metric” for bloggers to identify the bloggers he believes are most relevant in the space. Then he plots them with a size that shows how well-linked a blog is, and uses a physics model to cluster based on linkage.

Kelly then uses “attentive clustering” to color the graph – people who link to the same sources are colored the same way. There’s a clear cluster around conservative politics, and a visible cluster that’s conservative, pro-Israel. There’s a fringe group he calls “Islam critics”. On the other side, he sees clusters of progressive insiders, progressive outsiders, and progressive media critics. Other clusters are apolitical – economics, law, education, health and healthcare. Web cultures – Gizmodo, Make magazine – are also represented in the map. And there’s a cluster of journalism criticism, which Kelly notes is uncomfortably close to people who watch celebrities.

He characterizes the progressive critics as reasonably well connected to other conversations, and the conservative conversation as largely separate. Unsupriringly, a site like Newsbusters.org gets lots of attention from the conservative cluster… but does get some links from the big dogs on the progressive side. Factcheck.org is the mirror image – the big conservatives, and most people in the progressive space. Politifact is similar. Media Matters is further out towards the progressive fringe, though gets attention from big conservatives. Politicalcorrection.org is even further left.

MRC.org is mainly linked from the right, but gets good response from the journalism commentary cluster. Washington Post’s Factchecker blog gets equal attention from the left and the right, but lots of love from the journalists. CJR is loved by the left and the journalists, and invisible to the right. Sunlight Foundation has lots of traction in the tech community and is stronger thre than in political circles. For a comparison, Kelly offers snopes.org, which seems to be equally noted across the board.

Healthnewsreview.org, a site that focuses on corrections in the health and healthcare space, has excellent traction in one space, but almost no influence in other parts of the mediasphere. This offers some interesting implications for niche communication strategies, but offers some worries about information crossing from subject domains into the main conversation.

Kelly graphs 1000 top sites in terms of who links to them. The graph has two dimensions: left/right and political versus mainstream. The political fact sites range from the left to the right, but are strongly linked to by political sites. Some odd exceptions – CJR is left and fairly mainstream, while NPR is quite central and fairly mainstream.

Morningside has also looked, though in less depth, at a set of Twitter accounts that follow fact checking organizations. They picked a set of key fact checking twitter feeds and grabbed all of their followers. They looked for linkage and clustering and used k-core analysis to choose a densely connected set. What results is a space where conservatives appear to follow political fact checking more closely than progressives. (I’m not entirely clear on how Kelly is determining left-right within this set – I assume he’s hand-checking the clusters that emerge in his analysis, which is his standard operating method.)

Even a highly partisan site like politicancorrection.org has substantial followership from the right. Kelly drills down and sees clusters of followers in the Occupy movement, in the union and labor space, and in the eco/green space, as well as beltway insiders and people who study media. But he also sees a cluster of followers of conservative politicians, and a cluster around conservative media personalities.

How might we explain this? It could be that Twitter is where conservatives are making their stand in social media. Conservatives may be watching Twitter very closely and responding to each of these fact check interventions. It’s hard to know, though, as Kelly notes that Twitter is a space of “non-authentic actors”, both automated bots and coordinated groups of humans.