Analyzing and Visualizing Log Files: A Computational Science of Usability

Researchers in human-computer interactions know that software can easily be instrumented to create a trace of user events in the interface (which we call a log file) for later analysis. Using these data for studying usability or other HCI questions is certainly attractive. The data are cheap - data gathering is totally automatic. The data are gathered discretely, so the possibility of a Hawthorne effect is diminished. Data can be collected outside the laboratory, while the user is engaged in real tasks, so the results have greater ecological validity. The problem has always been what to do with the volume of data that gets generated. With so much data, it's difficult to determine what's useful and how to display the useful portion in a meaningful way