Thank you for posting, Tim. I could have this conversation over countless cups of coffee.

The other element Google and Cutts nailed is that of visibility. Sure, they weren’t telling the world precisely how their algorithm worked…

The other element Google and Cutts nailed is that of visibility. Sure, they weren’t telling the world precisely how their algorithm worked, but they provided details regarding changes, explanations for them, and much documentation and coverage of their implimentation. Further, this visibility was built into the product in a way Facebook hasn’t built its feed: search any topic and Google’s choices are on display. (This is getting muddied by further personalization, but that is a topic for another time.)

Facebook, by comparison, is a black box. Why is this article trending? What are you not seeing in your feed? How many people saw an article and how did they react? Filter bubbles are invisible to those inside. Quantifying and qualifying this effect, both for media outlets and individuals, would be an appreciated step.

For those who made it this far, I’d like to recommend a few books related to the ideas above which will spur new ideas, help organize your discussions, and (hopefully) help us craft new solutions:

The Master Switch, by Tim Wu: A strong argument for why information and communication companies should be held to higher standards, with invaluble historical context. Wu’s arguments and examples strongly support’s O’Reilly’s position that Facebook needs to take responsibility for the content and ideas they filter.
Nudge, by Richard Thaler and Cass Sunstein: An exploration of design and how it can influence our decisions without limiting them. A great way to dive deeper into the ideas behind O’Reilly’s Google example.
The Signal and the Noise, by Nate Silver. Numerous case studies in support of a thesis that machines coupled with human judgement always beat machines on their own. If there is a point where I disagree with O’Reilly, it’s that you should abandon editors for engineers. I think you need both, as Silver argues better than I, especially since editors can help with identifying and judging “good results.” Otherwise, the feedback mechanisms the engineers will work with will likely be the current, easily quantified and valued retention, time spent, clicks, and revenues–all of which are largely value neutral.