Slashdot Mirror
The Demographics of Web Search
adaviel sends a link to work out of Yahoo Research indicating that demographics can help Web searches; e.g. a women searching for "wagner" probably wants the 18th-century German composer, while for men in the US "wagner" is a paint sprayer. The Yahoo researchers claim that by taking user demographics into account, "they managed to get the chosen link to appear as the top-ranked result 7 per cent more often than in the standard Yahoo search." New Scientist mentions this research and two other innovative adjuncts to current search practice: following the mouse cursor as a proxy for eye tracking, and taking back bearings on online criminals by studying the searches they make. (The latter raises disburbing privacy questions: would you want Google trolling through your search data? How about governments?)
Wagner was a 19th-century composer, not 18th.
> would you want Google trolling through your search data? How about governments?
Heck yes I want Google trolling through governments' search data.
(Yes, I'm being facetious, but still. That Wagner example is pretty awful.)
Yes, that's really what we need...
What next, a search result that depends on your religion? If you type "Origin of the Universe", you get articles about the Bible if the engine thinks you're Christian, and scientific material otherwise?
They need to understand there is little value in subjective data. Their results are already biased enough, they should take steps to fix that, not make it worse.
> Applying demographic data like this is a non-sequitur.
What would be useful is if I could choose to search from a different persons/demographic's point of view. Whether for ebay, amazon, google.
For example say I am looking for a gift for someone else. Or I am helping someone else search for stuff. Or I'm the sort of person who has rather different interests but with search keywords that overlap.
Same goes for reviews of restaurants/movies/etc. What I like, someone else may detest.
Lastly, it could also be interesting (and even beneficial) to be able to more easily see things from other people's point of view.
... this idea smacks of a tool that's trying to be *too* helpful, and ends up getting in the way. Kinda like the old microsoft paperclip. I went and turned off this function in google accounts when I realized that my search results were being shaped based on my history, since that partially defeats my expectations of how a search engine behaves, and degrades the utility, insofar as the utility (to me the user) is based on receiving an unbiased sampling of the matches. I'm also troubled by this trend in the way that google delivers their news offerings, it seems that the logical progression of this is that we will mostly only be exposed to material that fit our highly individualized pre-existing reality bubbles.
The search results are not just a regex matching. A modern search engine, like Google's, returns a ranked list of search results to you, and this ranking already has bias: the Pagerank algorithm sorts the results based on how popular the page is, as measured by the number of incoming links to that page. Of course, that is the general gyst of Pagerank as of the Google founders' research paper back in the late 1990s, and undoubtedly Google and other search engines have fine-tuned their algorithms since then to return "better" results to the user. But the point is still that there is already bias in the results.
Make no mistake that Google has not already thought of similar search result ranking algorithms similar to that posed in this Yahoo Research paper. The difference is that Google does not have a research arm like Yahoo, so they do not publish ideas like this. In hindsight, the Google founders were foolish to publish their Pagerank algorithm in the first place, but they were still at Stanford then.