The fact that as you read the web the web reads you has been obvious since Mark Zuckerberg was in short pants. The ability of cookies, algorithms and other software to discern, at light speed, your identity, desires, and intentions lies at the heart of the web’s attractiveness as a commercial medium, and pushing that ability forward is critical to the competitiveness of web firms, particularly those serving up search results and ads.
A patent issued to Google this week – and noted by Slashdot – sheds light on one area where the search giant is hoping to enhance its ability to “read the user.” The patent covers a set of techniques for, as the authors cumbersomely put it, “rendering advertisements with documents having one or more topics using user topic interest.” Google recognizes that web pages or other online documents often have many sections and include many topics. The resulting ambiguity about what a viewer is looking at or is interested in complicates the placement of relevant ads on the page. An ad may be tied to a section or topic that is of little interest to the viewer and thus fails to capture his attention.
What Google has patented is a way to decipher which page regions and topics the viewer is interested in based on the viewer’s behavior after he’s arrived at the page. (Typically, ads are served based on the user’s behavior before he arrives at a page.) The patent’s scope, in defining the elements of behavior, is broad, even open-ended: “Examples of user behavior include (a) cursor positioning, (b) cursor dwell time, (c) document item (e.g., link, control button, etc.) selection, (d) user eye direction relative to the document, (e) user facial expressions, (f) user expressions, and/or (g) express user input (e.g., increasing the volume of an audio segment), etc.” Any of these actions could be used to trigger the placement of an ad or the replacement of one ad with another, as illustrated in the following figure from the patent.
This is yet another example of a patent that appears to cover a very broad range of analytical techniques. More interesting, though, is how it points to an expansion in the ways web companies will be able to, in effect, read our minds. Should the Google methods be implemented – and I don’t doubt they will, in some form – our computers won’t just feed back to companies information about what we click on or where we move our cursors but also, thanks to the ubiquitous webcam, information about our “facial expressions” as we peruse a page. We all have our tell-tales, the tics that give us away, and once you digitize them and run them through an algorithm, they’re pure gold.