Bo Begole

The routines of information work are commonplace yet difficult to characterize. Although cognitive models have successfully characterized routine tasks within which there is little variation, a large body of ethnomethodological research has identified the inherent nonroutineness of routines in information work. We argue that work does not fall into discrete classes of routine versus nonroutine; rather, task performance lies on a continuum of routineness, and routineness metrics are important to the understanding of workplace multitasking. In a study of 10 information workers shadowed for 3 whole working days each, we utilize the construct of working sphere to model projects/tasks as a network of humans and artifacts. Employing a statistical technique called T-pattern analysis, we derive measures of routineness from these real-world data. In terms of routineness, we show that information workers experience archetypes of working spheres. The results indicate that T-patterns of interactions with information and computational media are important indicators of facets of routineness and that these measures are correlated with workers' affective states. Our results are some of the first to demonstrate how regular temporal patterns of media interaction in tasks are related to stress. These results suggest that designs of systems to facilitate so-called routine work should consider the degrees to which a person's working spheres fall along varying facets of routineness.

This paper introduces a new representation for describing routine tasks, called temporal task footprints. Routines are characterized by their temporal regularity or rhythm. Temporal pattern analysis (T-patterns) can be used to isolate frequent recurrent patterns in routine tasks that appear repeatedly in the same temporal configuration. Using tf-idf statistics, each task can then be defined in terms of its temporal task footprint, a ranked list of temporal patterns along with their typical frequencies. Experimental evaluations using data of 29 days observing and logging 10 subjects showed that temporal task footprints of application windows, email and document usage outperform decision tree and SVMs in recognizing the subjects' tasks.

We present a method to infer the orientation of mobile device carried in a pocket from the acceleration signal acquired when the user is walking. Whereas previous work has shown how to determine the the orientation in the vertical plane (angle towards earth gravity), we demonstrate how to compute the orientation within the horizontal plane. To validate our method we compare the output of our method with GPS heading information when walking in a straight line. On a total of 16 different orientations and traces we have a mean difference of 5 degrees with 2.5 degrees standard deviation.

We describe the use of a statistical technique called T-pattern analysis to derive and characterize the routineness of tasks. T-patterns provide significant advantages over traditional sequence analyses by incorporating time. A T-pattern is characterized by a significant time window (critical interval) that describes the duration of this pattern. Our analysis is based on data collected from shadowing 10 knowledge workers over a total of 29 entire work days. We report on the statistics of detected T-patterns and derived correlations with participant perceptions of workload, autonomy, and productivity.