AAAI ‘06 (from Jen)
by Jen Golbeck
As Jordan posted, he and I presented our paper yesterday. I think it went very well. We had lots of interested questions at the end of the talk and after the session.
In addition to our session, I visited some others. I was particularly interested in the
Intelligent User Interfaces session, which included a paper co-authored by Cai Ziegler (one of my co-authors). That was a Nectar paper discussing the work in the GroupLens group at Minnesota. They reiteraeted a point that I had read in their work before: that accuracy as the measure of the quality of a recommender system is not the best way to go, and may even be harmful. Amazon’s recommendations are a good example of this - if you buy one book by Stephen King, it will recommend all of Stepehn King’s other books to you. You may read those books, but it’s not a very insightful recommender. Similarly, it’s not so interesting to look at what you’ve done and have a system that can predict what you’ve done. The more interesting thing is to make good recommendations about what the user might like in the future. Diversity and other measures that are not just about accuracy are also important. I love this point and totally agree.
Today there were also some good game theory talks, that took me back to the kinds of things I used to study before I came to mayland. Those talks were fun, and very smart.
