Measuring Privacy: Using Context to Expose Confounding Variables

Abstract

It is commonplace for those who support less restrictive privacy regulation on the collection and use of personal information to point out a paradox: in survey after survey, respondents express deep concern for privacy, oppose growing surveillance and data practices, and object to online tracking and behavioral advertising. Yet when confronted with actual choices involving the capture or exchange of information few people demonstrate restraint: we sign up for frequent flyer and frequent buyer programs, we are carefree in our use of social networks and mobile apps, and we blithely hop from one website to the next, filling out forms, providing feedback, and contributing ratings. Privacy skeptics suggest that actions should be considered a truer indicator than words. Even if people are honest in their positive valuation of privacy in surveys, in action and behavior they reveal even greater valuation of those benefits that might come at a privacy cost. In other words, people care about privacy, but not that much.The inconsistencies between survey responses and observed behaviors that skeptics gleefully observe require a nuanced interpretation – one that we have offered through our studies. We argue the disconnect between actions and survey findings is, not because people do not care about privacy but because individuals’ actions are finely modulated to contextual variables. Questions in surveys that do not fix such important contextual variables explicitly are highly ambiguous.A more nuanced view of privacy is able to explain away a great deal of what skeptics claim is a divergence of behavior from stated preference and opinion. People care about and value privacy – privacy defined as respecting the appropriate norms of information flow for a given context. When respondents are given a chance to offer more fine-grained judgments about specific information-sharing situations, these judgments are quite nuanced. This is problematic since public policy relies on survey measurements of privacy concerns – such as Westin’s measurement of individuals as privacy ‘pragmatists’ or ‘unconcerned’ – to drive privacy regulations. Specifically, Westin’s categories give credence to the regulation of privacy based by FIPPs, which relies heavily on assuring individuals notice and choice. We examine two historically influential measurements of privacy that have shaped discussion about public views and sentiments as well as practices, regulations, and policies: (1) surveys of individuals’ ratings of ‘sensitive’ information and (2) Alan Westin’s privacy categorization of individuals as fundamentalists, pragmatists, and unconcerned. In addition to replicating key components in these two survey streams, we used a factorial vignette survey to identify important contextual elements driving privacy expectations. A sample of 569 respondents rated how a series of vignettes, in which contextual elements of data recipient and data use had been systematically varied, met their privacy expectations. We find, first, that how well sensitive information meets privacy expectations is highly dependent on these contextual elements. Second, Westin’s privacy categories proved relatively unimportant in relation to contextual elements in privacy judgments. Even privacy ‘unconcerned’ respondents rated the vignettes as not meeting privacy expectations on average, and respondents across categories had a common vision of what constitutes a privacy violation. This study has important implications for pubic policy and research. For public policy, these results suggest that relying on one dimension – sensitive information, privacy categorization of respondent – is limiting. In particular, focusing on differences in privacy expectations across consumers obscures the common vision of what is appropriate use of information for consumers. This paper has significant public policy implications for the reliance on consumer choice as a necessary approach to accommodate consumer variance: our results suggest consumers agree as to the inappropriate use of information. Our study has called privacy concepts into question by showing ‘sensitivity’ of information and ‘concern’ about privacy are not stable in the face of confounding variables: privacy categories and sensitive labels prove to be highly influenced by the context and use of the situation. Our work demonstrates the importance of teasing out confounding variables in these historically influential studies.