Data Bias, Behavioral Choices, and Policy Nudge: Pitfalls and Opportunities in Urban Information Governance from a Behavioral Science Perspective
Abstract:
Recent progress of urban informatics provides new opportunities for more
refined and intelligent urban governance. The large amount of new data acquired
through Social Sensing contains rich information on urban dynamics and citizens’
behaviors, and implies potential for data mining and knowledge discovery. However,
observational bias in Social Sensing data is worthy of deep concerns. This paper, in
investigating the data bias problem, firstly enumerates various kinds of common new
data and their application scenarios in urban governance, and briefly discusses the main
properties of the three basic bias types, namely selection bias, information bias, and
counterfactual bias. Next, this paper analyzes the behavioral roots of the biases from the
perspective of behavioral science with an “exposure-cognition-action” framework, and
then proposes a methodological framework which utilizes natural experiment designbased causal inference to realize the decomposition of the pure effects of the three bias
types. Finally, this paper takes the typical data resources such as urban management
inspection records and law enforcement surveillance videos as examples, and
demonstrates ways of correction, attribution, and “nudge”-style utilization of data bias
with the proposed methodological framework which helps improve people’s well-being.