With this week’s paper we examine the connections between urban and trail technology use, considering how lessons from sensor-based science the woods affect urban sensing efforts.
- Dana Cuff, Mark Hansen, and Jerry Kang. 2008. Urban sensing: Out of the woods. Communications of the ACM 51, 3 (March 2008), 24-33.
- Brief revisit of reading group and introductions of anyone new
- Attendance: 4 people (1 professor, 2 graduate students, 1 undergraduate student)
- Summarize papers
- Discuss papers
This paper, authored by an architect/urban planner, a statistician, and a lawyer, outlined challenges regarding sensing efforts in urban areas. They began by describing successes in sensor-based science in the woods, where air, water, soil, etc. sensors could collect data continually without objection from the birds or worms. They contrasted that with efforts to do sensing in urban areas, where there are legal issues (people object to being sensed more than birds) and a greater likelihood of junk data and garbage analysis.
The issues from this paper resonated with the people in the discussion group, particularly in reflecting on past papers. Several people noted parallels with the challenges encountered with the Dix data, for which even a careful scientist and a conscientious science community generated faulty data (e.g., GPS malfunctions, forgetting to turn on or off a sensor) or questionable analysis (e.g., is there a “happy” day, or is that the wrong unit of time to examine). Challenges stemming from large data sets, including human-generated data, certainly have grown in importance in the 8 years since this paper appeared.
The calls for action certainly seem prescient today. The call for more data commons efforts is seen in data sets like the Dix data, government datasets, and the NSF requirement that all proposals include a data plan that explains how data generated in the proposed work would be made available for others to analyze. The call for distributed citizen-initiated sensing is seen in efforts like Google traffic (Waves), Foursquare, bar tracking, Facebook check-ins, openstreet map, CMU bus schedule project (Zimmerman), and lots of other crowdsourcing efforts.
And it was encouraging that the authors noted that user interfaces are both important and hard. Again, we saw that in the Dix data, which was made available but was reformatted and cleaned up multiple times–and gaining insights from it is still hard: McCrickard’s class had difficulty identifying interesting findings for a week-long homework assignment.
There were lots of ways forward that emerged from the discussion and paper. Many were related to the difficulty in knowing the right question to ask about data sets? A bottom up examination reveals trends and top down reveals questions, but there’s a danger in just finding what you’re looking for without a scientifically rigorous approach. The paper seems prescient 8 years later, and worthy of consideration moving forward.