- SSMS 1009
As the volume of user-provided content on the web increases rapidly, we rely more on automated methods for finding relevant, trustworthy and useful information in a timely manner. Traditional mechanisms for tackling this information overload problem include active techniques such as entering a search query over an index, and passive techniques such as receiving automated item recommendations such as with products on Amazon.com, or the highly personalized advertisements that we see on Facebook. Social information filters are the focus of this talk, particularly the fact that most of these filtering mechanisms are 'black boxes' that simply provide a final prediction or recommendation based on an initial set of input parameters. The majority of automated recommendation systems do not attempt to explain to the user how personalization is achieved, which can be annoying in many cases, for example, 'why is Facebook recommending me a face-lift'? In addition, personalization systems usually operate on user profiles gathered over time, meaning they can be outdated or irrelevant for a variety of other reasons. In this talk I will discuss potential solutions to the above problems and demonstrate on Facebook and Wikipedia how an interactive interface can increase both transparency of an information filtering system and users' satisfaction with recommended content.