Visualizations of Task Performance

CogTool is system built for user interface designers and usability specialists to quickly create interface mockups and then automatically generate Keystroke Level Models that provide accurate predictions of expert performance using the proposed designs. Prior to my involvement, the output of CogTool was simply a predicted time. This provides little or no design guidance to user interface developers. In response to this problem, I designed and implemented an interactive visualization that lets designers explore exactly what the cognitive model did to produce its performance prediction, and therefore gives additional clues as to what specific components of the task are inefficient. This visualization has been highly successful for both usability analysts and cognitive modeling researchers, and efforts are currently underway to allow other systems to plug in to CogTool and use the same visualization interface. CogTool is written in Java using the SWT toolkit and is open source.

Gesture-Based Media Playback

MaestroMedia is a fullscreen media playback system for the Macintosh inspired by the concept videos Nintendo released for the Wii video game system in fall of 2006. Built more than a year before the Wii was released, MaestroMedia provides an intuitive, gesture-driven interface based on a gyroscopic mouse. The system can recognize a combination of continuous and discrete gestures, with a user experience inspired by the iPod. In a sense, MaestroMedia is like a magic wand for your television, or a remote control with no buttons. MaestroMedia is written in Objective-C using the Cocoa API.

Location Finder for Windows Mobile

PeopleFinder is part of a larger research project underway at CMU to investigate how well people are able to specify privacy policies to govern information disclosure, and to develop user interfaces to help people develop policies that reflect their true disclosure preferences. I built a system for Windows Mobile phones that uses Intel's POLS software to determine a user's latitude and longitude by analyzing relative signal strengths from cell towers. The client uses push technology to request location information from other clients, and can visualize their locations on an interactive Google map using either satellite or map imagery. The system works over the phones' data connection using a web server as an intermediary, but can also be entirely peer-to-peer, using SMS as the transport protocol. PeopleFinder is written in C# for Windows Mobile 2005.