Yahoo’s man+machine algorithm: the numbers are in
Julia Kamin / August 3rd, 2011 / Tweet
So how well does personalization work, anyway?
Over at Yahoo, according to FastCompany, quite well. Since setting up their crack personalization team in 2009, clicks on Yahoo’s “Today” box have increased 270%.
That’s saying personalization makes us four more times likely to click on a link. Whether you believe personalization makes the internet more efficient, more fractured or more mind-numbing, that’s a pretty impressive number.
For those concerned about the self-looping and fragmenting effects of the filter bubble, the good news is that Yahoo’s algorithm is not entirely human-free. Editors are in charge of curating the 50-100 versions of the “Today” module that could pop up on your Yahoo home page; the bots just guide them to which stories work best and, ultimately, which take on “Today” you’ll see.
Humans are also, thankfully, still in charge of deciding when civics trumps the bottom line:
On the day Fast Company visited, President Obama was slated to give an important speech that evening on the draw down of troops in Afghanistan.The algorithm predicted that the story on the speech would do miserably with Yahoo visitors. And indeed, according to the dashboard, it wasn’t getting many takers. But the editors still flipped the override switch, ruling that the story would be shown to all visitors to the home page at least once, irrespective of what the algorithm said. It was, and Yahoo willingly took the hit on clicks. Some stories, the editors say, everyone simply needs to see.
I would love to see the same analysis Eli & Co. are applying to social networks aimed at examining the effects of the congressional redistricting schemes of the 1990′s which sought, laudably, to create more “majority minority” congressional districts. They succeeded, of course, but some argue the unintended consequence was a hard turn away from anything approaching a national consensus on important issues.
This phenomenon is not unlike personalization on the web: reduce my political choices based on certain demographic criteria – namely the racial makeup of my neighborhood, or its predominant party affiliation – and suddenly everyone who votes at my polling place thinks much like I do, and so do the candidates. The result? Candidates that appeal to narrower and narrower constituencies, unmotivated by a need to reconcile divergent opinions.
As Madison famously wrote in Federalist #10:
“…as each representative will be chosen by a greater number of citizens in the large than in the small republic, it will be more difficult for unworthy candidates to practice with success the vicious arts by which elections are too often carried; and the suffrages of the people being more free, will be more likely to centre in men who possess the most attractive merit and the most diffusive and established characters.”
In this case Madison was arguing for a larger electorate as the best way to dilute the power of one special interest, or an oppressive local majority, in favor of policies that favor the “public weal.” But his reasoning applies equally to the creation of “electoral bubbles” wherein an oppressive political minority (the Tea Party, e.g.) is able to wield disproportionate power because of the lack of electoral consequences for the representatives it elects. As race and ethnicity become increasingly fluid concepts in this country, progressives should look hard at whether real damage has been done to our public life and our democracy by these policies, and reconsider their usefulness. This may have been a necessary phase in our evolution, but if we want to reduce partisan deadlock in our government, the time may soon come when we must re-diversify our congressional districts or face the civic consequences of this political equivalent to the filter bubble.