Monday afternoon I was graciously invited by Andreas Weigend to be a guest lecturer to his graduate course on Data Mining and Recommendation systems. In general, the course evaluates the use of online personas and information to provide better user experiences and marketing.
It is readily apparent the increased penetration of mobile devices on everyone's lives. In the United States the iPhone has revolutionized how people daily interact with online information - using Wikipedia to investigate local history and marine reports to understand the freshness of fish at the market. In Asia they've long had the capability to find nearby friends or potential mates, point at buildings to query reality, and instantly send live media to one another. In Africa mobile devices serve as the primary infrastructure for communications, payment systems, and even voting.
With people continually carrying both a personal sensor and data device it's possible to glean powerful insights into behavior, desire, and action. Users are actively seeking to better engage with their surroundings and community.
The question is, how can we access this latent information in order to understand the individuals and offer them appropriate, and appropriately delivered, information depending on context. It is important to know if a user is currently traveling in a car to not interrupt them but provide information that is localized based on their mode of transit and trajectory.
A restaurant search is more valuable when it shows relevant suggestions 2 or 3 miles ahead then perhaps any restaurant that is 0.5 miles behind the car. Similarly, a pedestrian has no use for a search result that is over a mile away, but is willing to patronize within several blocks - but also depending on perceived safety and familiarity with the various regions.
The lecture was a survey of both the abstract concepts of mobile geolocation, proximity, trajectory and data mining, as well as examples of emergent technological and anthropological solutions. Flickr serves as a very coherent example of providing a hoard of information on user behaviors (time, content), lifestyle (pictures of children, parties, travel, business), and obviously location.
Services could identify young fathers of children that travel often and afford expensive cameras to offer child-care services, luxury family sedans, or maybe even high-end strollers (I'm looking at you Bugaboo).
And the techniques don't apply to purely marketing needs. It would also be possible to use contextual relevance and behavior to understand the flow of a city for better planning. Of course, I could imagine more nefarious purposes as well as governments seek to track and prevent dissident movements.
The utilization and integration of mobile devices, or more generally ubiquitous computing, has yet to mature - but the platforms for experimentation and demonstration are now common and the general public is becoming more comfortable with at least the understanding of location-contextual relevance.