Influence Networks is an open-source, collaborative directory of relationships between people, institutions and companies. Each relation has its own level of trustworthiness, so that facts can be distinguished from noise.
En charge du pôle datajournalisme chez OWNI, je travaille avec les designers, les developpeurs et les journalistes pour produire des applications de journalisme augmenté de données et de code, mettre en place des actions de crowdsourcing et des serious games.
In October, 2010, during the Personal Democracy Forum in Barcelona, several investigative journalists explained how they managed to uncover corruption using network analysis. One of them, Dejan Milovac, wrote a story about a construction project on the Montenegrin coastline. He deconstructed the financial networks around the resort, and showed how local politicians were involved in an enterprise that was ostensibly going against all environmental rules. Below is the image illustrating result of the investigation:
This diagram holds some margin for improvement, beginning with readability. What’s more, the relationships exposed in this investigation could be useful to other journalists working on similar subjects. As such, reusing Milovac’s work would be a daunting task.
Network analysis has become a popular topic in several newsrooms. Channel 4 produced Who Knows Who, a database of relationships linking British personalities. In Hong Kong, the South China Morning Post launched Who Runs HK?, a similar project. These interfaces, although run by journalists, remain closed, and cannot be linked to open formats.
On the geekier side, Little Sis is another database of relations. It’s collaborative, open and has its own API. 57,000 people appear in there, with close to 300,000 connections. The only problem that remains is the bits of information in Little Sis are not validated and only an alert mechanism (flagging) allows for fighting disinformation. Given the sensibility of such a project, it is very likely that lobbyists will take over and, at some point, game the system.
There was a need for a network analysis tool who could be used by journalists and that be both open and reliable. Influence Networks aims at solving this problem. The platform allows for anyone to add a ‘relationship’ in the database (though the “add a relation” tab). The bit of information the user inputs is given a “rumor” status as long as its reliability has not been assessed by another user.
The “review a relation” tab allows for vetting the credibility of relations added by others. The user who validates can rate the relation on a scale going from “rumor” to “established fact.” The rating the relation receives also depends on the trust level of the user who validates it.
Let’s take an example. Mathias registers on the platform. He starts with a trust level of 1 out of 5. He adds a valid piece of information, complete with source. The information is validated by a user who has a trust level of 5 out of 5. The information is then given the status of “established fact” and Mathias’ trust level increases by 0.5.
Mathias then adds another bit of valid information, but the information is validated by George, who has a trust level of only 1. This time, since we have no way of knowing about George’s trustworthiness, the information’s status is upped just above “rumor”, but not much more.
As of today, right after launch, the database contains only a few ‘relationships’. We will build a batch-add feature in the coming months, so that users can, for instance, upload a meeting’s attendance list. What’s more, the entities (people and organizations) that can be described in Influence Networks all come from Freebase, which is like a machine-readable Wikipedia. That means that if the entity you are looking for is not featured on Influence Networks, you can go to Freebase and add it there. In the coming months, we will add a feature to do just that within the Influence Networks interface.
Influence Networks’ aim is to allow any journalist or citizen to call for a collaborative investigation on a given topic. Let’s say you want to check on all contractors in the US military. You could do that by hand (very time consuming) or ask for contributions that you then check by hand (again, a very long process).
With Influence Networks, all data is collaboratively validated and managed on a machine-readable format. That means that data can be linked with other relevant repositories and speak for itself. A search between “US government” and “military contractors” could reveal who might have a conflict of interest when going to war, for instance. Remember in late 2001, when KBR, a Halliburton subsidiary, was granted a 10-year contract wherever US troops went? A network analysis tool at the time might have made the ties between KBR and vice-president Cheney clearer.
As a matter of course, the tool is not meant for investigations based on secret documents. However, it does structure and reorganize publicly available information. After all, open-source intelligence is a fast-growing field – in journalism as well.
Influence Networks was born out of a collaboration between OWNI, Transparency International, Zeit Online and Obsweb (Metz University). This group presented the project to two international competitions in innovative journalism. We did not make it to the final round of the Knight News Challenge but did make it to the top 75 (out of 1,500+ candidates).
We are, however, part of the 10 finalists in the Uutisraivaaja Challenge, a similar contest organized by the Finnish Sanomat Foundation. The €10,000 grant they awarded us will allow for the development of several other features and for continuing to look for resources to take the project to the next stage.
The source code of the application is open under an MIT license. Don’t hesitate to examine what’s going on under the hood over at GitHub and contribute to new features!