Akkam’s Razor

We knew it was inevitable - people are doing funny, cool, and strange things with the AOL datadump.  We have a browstable interface at the purposefully named aolsearchdatabase.com.  There's discussion in some corners of contests for finding the most deranged AOL users, AOL query poetry, and short stories.  I'd fullly expect by the end of day to have some pretty data visualizations, related keywords, image representation of queries, destination URLs, and who knows what else.  How about a version of jeopardy where you get the results and have to guess the query?  What a great way to teach effective query formulation!

The voyeuristic aspect of watching online web activity can be addicting - check out the various spy services that allow you to watch online behavior, such as those for digg (spy, swarm, and stack), del.icio.us, and other.

As far as the database is concerned, there's a few interesting things.  For starters, it only includes persons using the AOL client, as opposed to someone searching directly from AOL.com.  It may be possible to make some inference as far as to the users tech-saviness since they wouldn't go to a browser instead.  Also, several queries have been people "searching" for a URL…and advanced user might do that to see if a URL is mentioned somewhere, but in this case, it's probably a lack of web experience.

So, we already know that there's plenty of real identifiable information that people willingly searched for on the web, with the assumption that those searches wouldn't end up on the web, such as SSNs, phone numbers, employers, names, and email addresses.  By reading these queries, and adding in your own perception and biases, you can form your own self image of who the user is, what there doing, and how their feeling.  But because of those perceptions and biases, there are also some false-positives, which is why datamining to stop crime or terror is a bad idea (because it takes resources to investigate every finding, and every finding is true unless proven otherwise), whereas the use of datamining for "good" - such as "honest marketing" - presenting people with products and options that they are genuinely interested in - can be less costly and more beneficial.