Social Characteristics
Summary
This article proposes a framework for the design of privacy experiments in a social network setting. The framework can also be used as a method to evaluate experiments from multiple privacy contexts (retail, health, etc.) in order to apply their methodology or findings to social network privacy studies.
This is an academic style paper, about 15 pages long. View Full Article as pdf.
Introduction
Privacy, a term rooted in ethics by definition, describes a zone or status free from public view and attention. Privacy is a sanctuary that overlaps with the perception of self so heavily that it can be considered a right, protected by policy and social norms, though it lacks a firm interpretation. Experiments to define privacy in the online domain incorporate a diverse set of characteristics which makes interpretation of study findings between and among privacy experiments difficult. The intent of this paper is to introduce a framework for evaluation and design of privacy studies with a focus on social network sites. A list of characteristics were elicited from non-social network privacy studies and compared to social network studies to identify a manageable, relevant set. These characteristics were identified as the data recipient, risk of context collapse, and disclosure intent. For each characteristic, two to three values were identified and organized into the proposed framework. This paper defines these characteristics and values and their impact on design, as well as potential ethical risks.
Background of Privacy Experimentation
In the digital age, the privacy can be described as the boundary for the free exchange of information pertaining to an individual or group. Determining this boundary in its many contexts is the first step to defining ethical practices and policies for data collection and handling. However, the contexts in which a definition of “reasonable” or “pragmatic” privacy exists depends on a myriad of dimensions including comprehension, potential harm, risk calculation, the mechanism of collection, etc., resulting in stagnation for online privacy policy development. Fortunately, the sustained growth of the internet economy and the current political environment have brought efforts to formalize privacy from the ruminations of sociology and psychology to the forefront of ethical debate and data science experimentation. While a cohesive interpretation of online privacy norms may be an unattainable goal, even a rudimentary attempt is necessary for effective policy development and implementation.
The complexity of the issue lends itself not only to experimentation, but to data science specifically in order to expand the breadth of analysis across different populations and online functions. However, given the plethora of established online mechanisms for information transfer and with the diversity of information types, the variety of experimental design offers a tangled definition of online privacy. Correlating these studies and their variables requires a web of interdependencies and assumptions. Some have distilled these into frameworks of core characteristics. These frameworks are critical to interpreting findings from multiple domains and contexts, in hopes of developing a cohesive definition of privacy.
Currently, scientists seeking to characterize privacy have methods such as a taxonomy and analytic tools. Solove’s taxonomy focuses on the life cycle of information in the general and digital environment, specifying opportunities for harm or damage (2006), while Mulligan, Koopman, and Doty’s analytic tool strives to clarify the function of privacy and its value by evaluating design and guiding debate using a set of dimensions (2016). These methods have been incorporated into applications to evalua
te privacy risks in social networks already by assuming some dimensions as fixed and specifying some dimensions as variables. For example Liu and Terzi (2010) utilize sensitivity and visibility in an equation to evaluate the privacy score of members in a social network. The sensitivity variable reference the dimensions of harm as well as Solove’s processing activities (Mulligan et al., 2016, and Solove, 2006). The visibility variable incorporates dimensions of scope and dissemination risk, while assuming collection as a constant (Mulligan et al., 2016, and Solove, 2006). Using these variables to calculate a user’s privacy score, Liu and Terzi (2010) have proposed a social-media specific privacy tool for future application development, education, and experimentation. This tool claims to be the first calculation of privacy risk for social network users. However, general online privacy has an expansive experimental repertoire, which could be incorporated into social network privacy analysis using the framework below.
Framework Visual- from Conclusion
