Page 450 -

14.5 Summary 441

Like physiological data discussed in Chapter 13, sensor-based ubiquitous com-
puting also requires thoughtful planning and careful analysis including preprocess-
ing, filtering, detection of specific types of signals, classification of activities, and
storage and management of data—possibly both raw and processed and related pat-
terns. Researchers have proposed a variety of processing pipelines (Ghose et al.,
2013), data monitoring tools (Bannach et al., 2010), reference datasets (Reiss and
Stricker, 2012), data integration strategies (Schuldhaus et al., 2014), and database
architectures, including the use of so-called NoSQL databases (Zeni et al., 2014), to
address these challenges. Details of these components will generally be dependent
upon sensor capabilities, requirements for data storage, and the specific questions
being asked.
Triangulated coordination of data collection and interpretation can often be
highly informative. A study of movement throughout places in the home illustrates
the potential for coordination of automated data collection with qualitative data—in
this case, interviews (Aipperspach et al., 2006). To understand the patterns of activ-
ity in homes, researchers placed location sensors at various points throughout several
homes. These sensors captured where people were, when they were there, and for
how long. Mathematical models were used to combine individual events in the log
files into meaningful aggregates that identified “places”—locations of significant
activity in the home. The models were evaluated by comparing the automatically
identified places with the results of interviews with the participants. Interviews with
the participants had the added advantage of providing context to explain some of the
results of the models. In one case, models identified a “place” that included both a
kitchen table and a living room couch. Interviews with the residents of this particular
home indicated that this data was collected during the course of a birthday party,
when they were continually moving between the kitchen and the living room, act-
ing as if the two locations were part of one larger space (Aipperspach et al., 2006).
Automated methods that focused only on the contents of the activity logs would not
have had access to this more nuanced explanation of resident activity.

14.5 SUMMARY
As the growth of the Internet and the availability of low-cost sensors led information
and computing into ubiquitous and familiar roles pervading all aspects of everyday
life, it seems only natural that these technologies would play key roles in HCI re-
search. Remote usability studies simplify the process of conducting usability studies
while providing access to larger scales of data. Human computation systems have
opened the door to an entirely new type of data collection, harnessing the power of
networks to engage many individuals in completing small tasks providing insight
unavailable through computational tools. Sensor-based systems allow for the easy
collection of new types of data in volumes not previously imaginable.
Online activity also provides a rich source of data for close examination of com-
plex interactions and communication patterns. Examination of online discussions