Facebook’s latest influence study is out, and the conclusions are not terribly surprising. You share information that your close friends share, but also things your not-so-close friends (or, your ‘distant contacts’, or ‘weak ties’, in network theory parlance) post. Thus, summaries of the study conclude, disproving the claim Facebook is an ‘echo chamber’, a set of behaviors many have insinuated is eroding our society, ingraining us in our ways and making life poorer through depriving us of tough choices about what we believe.
This is already leaving aside a glaringly obvious element. People wouldn’t be friends, even on Facebook, with people they don’t already share large swathes of cultural and economic common ground with. I am not issued a standard set of normative friends upon arrival, that’s rebalanced periodically to ensure all global viewpoints are represented. Reasonably, if Facebook is my only touchpoint with weak tie Jane Connection, it doesn’t mean she’s at the complete opposite end of the social and ideological spectrum to me. Some commonality brought us together, and I’d argue that’s strong enough to lend a coloration to the information he shares and makes me already predisposed to accepting it.
But, I can’t enter into a lengthy analysis of the paper until I actually read it. For now, more interesting matters.
The brilliant and able data scientists at Facebook have an unique porthole into some of the most amazing and interesting behaviors in human history. They’re able to observe major elements in how we fall in love, how we break up, how we celebrate birth and how we mourn death. They are able to judge very interesting things about human nature from these things. But, one must assume, their aspects of inquiry into the human condition are tempered by the desire of its executives to prove out Facebook’s advertising model, and the ability of Facebook to further monetize these events (or, the more prosaic ones, like when we mention our love for Starbucks or a positive experience at Hertz Rent-a-Car). Facebook actively works with advertising analysts to refine the products it sells marketers, so it should likely continue to do so more intensely as it grows.
Facebook is also constantly changing features in its service. Its EdgeRank algorithm, which determines what you see in your News Feed, is similar to Google’s PageRank, and a coveted position for marketers. If you’re a brand, even if millions of people have clicked ‘Like’, your content, which you may have spent millions of dollars to produce, won’t be seen by any of those millions unless someone engages with it, by Liking or commenting. If it’s not interesting, it won’t be seen. The more it’s interesting, the more it’s seen.
Trouble is, EdgeRank is largely a black box. Facebook’s Preferred Developers presumably have an inside edge, or at least a cobbled-together set of metrics with which they can determine how quickly something will take off.
But again, I’m straying from the point. The point is this: Facebook’s data studies should be assumed to be fundamentally serving Facebook’s interests. If it came to conclusions otherwise, why would it be released? Further, many of the statistics around behaviors on the web are commissioned and carried out by companies with vested interests in promoting the data. Security companies publish data on teenage hackers, for instance, or online persona management companies publish data on the proliferation of online personas. ‘These behaviors exists, so should we’ is communicated.
This is why I propose the Shining Volunteer Facebook Botnet of Truth and Victory to lead the way to transparent algorithm documentation.
It’s as simple as this: you sign away access to a moderately omnibenificient force that can monitor your news feed and occasionally post test elements, monitored by others in neighboring networks. Presumably it wouldn’t take more than a small percentage of groups to be able to make meaningful conclusions about the way EdgeRank works. Major changes would provoke an algorithm report to show what’s different. Maybe it would show that Coca-Cola’s content is altogether 10 times more important than Tiny Brand X’s content.1
This is a similar proposition to the idea of counter-algos in the high-frequency trading world, algorithms that try to out-act their counterparts. But this one acts on behalf of users of a system rather than its owners. The analogy that comes to my mind is that of a river and a dam. A dam may be owned and operated by a power company, used to generate power. But the water and the river are public property, and the department of the interior monitors the water level, and the releases from the dam, constantly, keeping track of flows and temperatures for recreation and the health of aquatic life. In the case of monitoring the health of our information flow, though, we need to actively allow some force to pretend to be us for a few moments to stick its toe in the water.
- I’m not a conspiracy theorist when I imagine brands that spend $10x more than others have some sort of advantage in EdgeRank. This would make good business sense for Facebook, rewarding those that buy comprehensive display packages with a leg up on those that can only afford to create compelling content. [↩]