Interrupted Interfaces

Have you ever watched a person use a mobile phone?  It’s difficult because mobile phones are, well, mobile. People are not still when talking on a mobile phone:  they walk, drive, shop, abruptly leave meetings, and so forth.  Mobile phone users move, and as a result, usually do something else at the same time. 

It is difficult to do two things at once. In fact we usually don’t. Research conducted at NASA found that when people appear to be doing two things at once, they really are rapidly switching back and forth between tasks. This is certainly true of mobile phone users. Researchers at the University of Finland found that most mobile phone use happens in 4-second bursts. You look up a name on your phone, look where you are walking, then dial the number. Trying to do two things at once turns a normal interface into an interrupted interface.

Multi-tasking while using an interface is common: drivers use a GPS; doctors use an electronic health record while listening to a patient; call center representatives look up information while talking to a caller. In all these settings, the demands of multi-tasking transform an otherwise normal interface into an interrupted interface.

How do you evaluate the usability of an interrupted interface?  A typical usability test, where recruited novices perform prescribed tasks, won’t be helpful. Using systems in dual-task settings requires experience and expertise not available to first-time users. Finding experienced users of a system is time-consuming, expensive, and in the case of systems not yet deployed, impossible.

At the recent UXMasterclass hosted by User Centric, a talk by Craig Miller of DePaul University described an alternative approach to interface evaluation that can be used with interrupted interfaces. Rather than observing novices, Miller described how the expected performance by experts on an interface can be modeled. The designer analyzes how a task would be performed on the interface, identifying each mouse selection and keyboard entry involved. Using a framework (GOMS) that has been used extensively over the past 25 years, a model of how an expert would perform these actions is constructed. Prior empirical research then is used to provide time estimates for each step. The result is an estimate of how long the task will take; it is a measure of the efficiency of the interface.

While efficiency is not the only dimension important to usability, for interrupted interfaces, it is the critical dimension. If users have only 4 seconds to accomplish a task, the interface has to be efficient.  When interruptions are commonplace, an inefficient interface quickly becomes an unused interface. Modeling the efficiency of interrupted interfaces will help ensure your interface is not only used, but is usable.