In Spring 2004, I have been working on Usability Testing using dual participants and on a census website test using pet rather than human data. The Usability Testing project involved asking two participants to discuss site navigation decisions with each other in order to assess whether that method produced similar results to doing a real-time interview. The pet census project was a proof-of-concept for the U.S. Census, to convince them that non-confidential data (I'm not going to steal your dog's identity.) could be used to test Census interfaces.
Both will be covered in forthcoming Tech Reports from the HCIL and LAPDP.
My primary research focus right now is on discovering navigation patterns in WWW search and tying them to individual difference measures of participants. There are several ways to approach this, including mathematically modeling the behavior and finding ways to visualize the behavior of actual participants. I intend to write simple mathematical models to describe this. I am also trying to devise a visualization which will allow me to see and share the patterns I discover.
The methodological approach I am using is one that was first put forth by another member of my advisor's lab about 16 years ago. It involves stripping the content from pages and only giving probability or relevance measures for the likelihood of finding a target in a particular branch of the web site (or in his case, file system). This approach was described in a 1989 tech report by Kent Norman and Scott Butler (CS-TR-2230, CAR-TR-432).
Hopefully, I can use this approach to write my Masters Thesis.