I spent Saturday at Columbia University, attending a mini-conference devoted to the subject of the disintegrating news industry and the future of the American public sphere. Despite never once getting called upon in 6 hours of hand-raising, and having to listen to some less-than-informed comparisons between the news industry and the music industry, it was a pretty interesting event, and I met some very smart and cool people.
One of the interesting ideas that came out of the conference was generated by me and Jack Balkin, in response to a talk by Bruno Latour about the negative effect of "attention deficit" on civic participation. In its essence, the idea is this:
THE PROBLEM:
- It's difficult for even a civically-engaged person to develop confident opinions about a broad range of political issues, mostly due to our limited attention and resources.
- This difficulty is exacerbated by (a) the dwindling resources of the professional journalism industry, and (b) the increasing [perceived] partisanship of the mainstream media, which reduce both salient information and the vital bond of trust between information-seekers and providers
- The result is increasing homogenization/polarizing of platforms, and civic disengagement
A SOLUTION:
- A recommendation engine for political facts/opinions/news that captures and reflects the underlying value systems that drive civic engagement by individuals, and delivers customized political information to those who seek it
The reason I cited Pandora specifically at the conference (when I finally stopped waiting to get called on and just started talking) is that I believe its tag-based database infrastructure, which charts the relationships between the intrinsic qualities of songs, rather than between their audiences, provides a promising alternative to more traditional collaborative-filtering recommendation techniques. Whether applied to music or politics, collaborative filtering tends to have an "echo chamber" effect, reinforcing the "tyranny of the majority" and supporting traditional generic categorizations. In politics, its effect would be to provide obvious recommendations that already match the reductionist platforms of the two major parties for most users. A Pandora-style architecture, on the other hand, would have the plasticity and nuance to deliver counter-intuitive news and information. Just as the music site now offers recommendations that fall outside of people's established generic preferences, a political version might offer Republican points-of-view to Democrats (or vice-versa) much more frequently, and accurately to the needs of individual users.
The problem with adopting the Pandora model whole-cloth would be its reliance on experts; its database is populated by a team of highly-specialized music analysts, who apply their expertise to the coding process. Clearly, to put a political recommendation engine in the hands of such "experts" would be problematic for several reasons -- the most important of which would be that they would provide the same reductionist coding as a collaborative-filtering engine, being steeped in the dominant Manichaean political framework. Much more useful (not to mention more robust, less expensive, more timely, etc) would be to crowdsource the database generation using a process similar to Wikipedia's.
So, in short, the political engine would use:
- Wikipedia-style data generation and management
- Pandora-style database architecture and processing
- Newsreader-style, queryable front-end..? (not so sure about this part yet)
The other benefit of this system would be that the degree of complexity introduced by processual firewall between data generation and database analysis would limit the potential for hacking, google-bombing, SEO, or whatever you want to call "tipping the electronic scales." In other words, if well-designed, it would be very difficult for a third party to pre-determine the opinion and data retrieved by any given individual -- just as Pandora is more or less immune to record label promotional efforts (they could, of course, accept payola and hack their own database, but founder Tim Westergren is vociferously against such meddling, even if it would save his business from going under).
Clearly, there's even more to it than this (the idea kind of sprung wholly-grown like Athena), but these are the rudiments. I've discussed it with a few smart people in the past 2 days, and everyone thinks I should try and build it... Comments? Suggestions?
Recent Comments