How the Yahoo homepage predicts your clicks

If you want details on the tech. go read some of John Langford's publications (<a href="http://research.yahoo.com/John_Langford" rel="nofollow">http://research.yahoo.com/John_Langford</a>) and the rest of the Machine Learning group at Yahoo (<a href="http://research.yahoo.com/Machine_Learning" rel="nofollow">http://research.yahoo.com/Machine_Learning</a>). I sometimes joke that Myna (<a href="http://www.mynaweb.com" rel="nofollow">http://www.mynaweb.com</a> -- ob plug for my startup!) should be renamed the John Langford Appreciation Society.

<p><pre><code> *"The system knows that women generally favor stories about Brad Pitt, but after some real-time analysis, it can quickly realize that men are far more like to click on a Brad Pitt story that involves a sports movie."* </code></pre> This is like Michael Frayn's first comic novel "The Tin Men" (1965), about programming a computer to produce an "automated newspaper".<p><pre><code> "But people really preferred an air crash. ... What people enjoyed most was about 70 dead, with some 20 survivors including children rescued after at least one night in open boats. They liked it to be backed up with a story about a middle-aged housewife who had been booked to fly aboard the plane but who had changed her mind at the last moment."</code></pre>

Yury Lifshits (then a research scientist at Yahoo! Research) did a talk at Estonian Summer School on Computer Science called "Intro to Content Optimization" [1], where he explained one of the approaches. It makes an interesting read.<p>[1] -- <a href="http://yury.name/esscass/" rel="nofollow">http://yury.name/esscass/</a>

very interesting to see how steadily machine learning algorithms creep into our news feeds (eg. facebook) and deliver us with the the links we're most likely to click. but i fear that this will lead to a serious decline in editorial quality, leaving an endless stream of cute kittens and hair styling tips. who will tell us what's really going on if some algorithm decides what's newsworthy and what's not?