The creators of technical infrastructure are under social and legal pressure to comply with expectations that can be difficult to translate into computational and business logics. This dissertation bridges this gap through three projects that focus on privacy engineering, information security, and data economics, respectively. These projects culminate in a new formal method for evaluating the strategic and tactical value of data: data games. This method relies on a core theoretical contribution building on the work of Shannon, Dretske, Pearl, Koller, and Nissenbaum: a definition of situated information flow as causal flow in the context of other causal relations and strategic choices.
The first project studies privacy engineering's use of Contextual Integrity theory (CI), which defines privacy as appropriate information flow according to norms specific to social contexts or spheres. Computer scientists using CI have innovated as they have implemented the theory and blended it with other traditions, such as context-aware computing. This survey examines computer science literature using Contextual Integrity and discovers, among other results, that technical and social platforms that span social contexts challenge CI's current commitment to normative social spheres. Sociotechnical situations can and do defy social expectations with cross-context clashes, and privacy engineering needs its normative theories to acknowledge and address this fact.
This concern inspires the second project, which addresses the problem of building computational systems that comply with data flow and security restrictions such as those required by law. Many privacy and data protection policies stipulate restrictions on the flow of information based on that information's original source. We formalize this concept of privacy as Origin Privacy. This formalization shows how information flow security can be represented using causal modeling. Causal modeling of information security leads to general theorems about the limits of privacy by design as well as a shared language for representing specific privacy concepts such as noninterference, differential privacy, and authorized disclosure.
The third project uses the causal modeling of information flow to address gaps in current theory of data economics. Like CI, privacy economics has focused on individual economic contexts and so has been unable to comprehend an information economy that relies on the flow of information across contexts. Data games, an adaptation of Multi-Agent Influence Diagrams for mechanism design, are used to model the well known economic contexts of principal-agent contracts and price differentiation as well as new contexts such as personalized expert services and data reuse. This work reveals that information flows are not goods but rather strategic resources, and that trade in information therefore involves market externalities.
