Year: 2023
Type: dissertation
Abstract: As computer systems handle transactions and make decisions about our lives, the opaqueness of such systems means that when faults occur, the negative impacts of these faults are often unfairly passed on to the individuals that are subject to these systems. \nThe aim of this thesis is to show how transparency enhancing technologies, technical mechanisms that support transparency about a system, can be used to address these issues. \nTo do this, this thesis provides an analysis of transparency enhancing technologies from their technical design to their practical usage. \n \nThe first part of this thesis systematizes transparency enhancing technologies, providing an analysis of the threat models and technical mechanisms that can support transparency. \n \nThe second part of this thesis gives an example of a transparency enhancing technology designed to provide transparency that enables publicly verifiable audits of the kind of access to data requests that are made by, for example, law enforcement and medical practitioners. \nThis work shows how distributed logs can be used to provide the infrastructure necessary for such audits and how publicly verifiable audits in the form of statistics (including multivariable statistics) can be produced with assurances that the privacy of individuals who relate to the information that is audited is respected. \n \nThe third part of this thesis discusses how transparency enhancing technologies can allow us to move beyond simply transparency and accountability. \nDrawing on the legal literature that has concerned itself with the legal effects of code as it enforces norms and the legitimacy of such effects, I discuss the ways in which code can be contested through the use of transparency enhancing technologies, in contrast to other kinds of accountability technologies based on compliance, and, because it is code itself, how transparency enhancing technologies can be legitimate, unlike other ways of contesting code.
Author: Hicks, Alexander
Institution: University College London
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7028003280
