Wednesday, August 10, 2005
trend to watch: usage studies
A very interesting recent development has been the marriage of device-use logging data with qualitative methods to develop a more complete and rich view of situated data. Earlier this year CHI held a panel, "Usage Analysis: Combining Logging and Qualitative Methods." Think of it as updating the unobtrusive measures approach pioneered by Eugene Webb some 40 years ago.
The approach brings grounding to observational methods. Observational methods are resource intensive, which means that "exploratory" observational research is often not efficient. Rather than looking at everything, suppose one could focus observation on specific issues that you know seem peculiar, and you would like to know more? Log data might offer guidance for focusing research.
Analyzing log data has become standard practice for web sites, but the possibilities are far broader. Server logs exist not just for web sites, but other server-hosted applications as well, which can be an interesting window into the behavior of remote users, especially mobile ones. Client-side logging would seem to generate too much data for sustained exploratory analysis. But one can choose to automatically log only certain events, if one is interested in looking at a particular behavior or application. Something as simple as a cell phone call registry contains potentially valuable information. In addition to server and client based information, the telecommunication network captures interesting usage information, such as who was called, duration of call, or data file sizes. Such billing information can be easily mined for patterns. The information from all these sources can reveal screen interaction data, user activity data (What applications does the person use? At what locations was the person?), and social interaction data (With whom does the user communicate?).
One powerful benefit of log data is that it is time-stamped. One gains insight into how people spend their time. One can also correlate different events, to examine how people juggle tasks, deal with interruptions, or work with colleagues. I can even imagine tying log data to other time-stamped information. Suppose you had a networked self-service kiosk in a public space. You notice from server logs unusually heavy usage at a certain time. You might be able to review closed circuit video of that same time to see how customers were dealing with situation.
As devices continue to gain functionality (GPS, video, etc.) the potential for log data will expand even more. Such data can provide a "big picture" view of behavior, and also provide a springboard for follow-up contextual research.
The approach brings grounding to observational methods. Observational methods are resource intensive, which means that "exploratory" observational research is often not efficient. Rather than looking at everything, suppose one could focus observation on specific issues that you know seem peculiar, and you would like to know more? Log data might offer guidance for focusing research.
Analyzing log data has become standard practice for web sites, but the possibilities are far broader. Server logs exist not just for web sites, but other server-hosted applications as well, which can be an interesting window into the behavior of remote users, especially mobile ones. Client-side logging would seem to generate too much data for sustained exploratory analysis. But one can choose to automatically log only certain events, if one is interested in looking at a particular behavior or application. Something as simple as a cell phone call registry contains potentially valuable information. In addition to server and client based information, the telecommunication network captures interesting usage information, such as who was called, duration of call, or data file sizes. Such billing information can be easily mined for patterns. The information from all these sources can reveal screen interaction data, user activity data (What applications does the person use? At what locations was the person?), and social interaction data (With whom does the user communicate?).
One powerful benefit of log data is that it is time-stamped. One gains insight into how people spend their time. One can also correlate different events, to examine how people juggle tasks, deal with interruptions, or work with colleagues. I can even imagine tying log data to other time-stamped information. Suppose you had a networked self-service kiosk in a public space. You notice from server logs unusually heavy usage at a certain time. You might be able to review closed circuit video of that same time to see how customers were dealing with situation.
As devices continue to gain functionality (GPS, video, etc.) the potential for log data will expand even more. Such data can provide a "big picture" view of behavior, and also provide a springboard for follow-up contextual research.
Link to this article. posted by Michael Andrews @ 8:27 AM 
