Helen Nissenbaum’s research while affiliated with Weill Cornell Medicine and other places

... AI inferences present significant challenges to conceptions of privacy, both in theory and in data protection practice. The privacy conceptions most susceptible to erosion by AI's inferential power are likely those grounded exclusively in categorical distinctionssuch as classifying data as sensitive or non-sensitive -while at the same time uncritically accepting AI-generated inferences as inherently valid [37]. Critical data scientists, particularly members of the FAccT community, have demonstrated the adverse impacts of invalid inferential models, especially due to biased misrepresentations that result in improper and unfair predictive descriptions of individuals they cannot understand, correct, or control (e.g., [36,38,49,51,98,105,106]). ...

Reference:

Privacy of Groups in Dense Street Imagery

... The question of values in games or sensitive design is approached in studies 5, 6, 8, 14, 15 and 16 with the tools "VASE", "Grow-A-Game", a curriculum for "Value Conscious Design", "The Values Cards" and "The Values at Play Framework" (Barendregt et al., 2021;Belman et al., 2011;Belman & Flanagan, 2010;Flanagan et al., 2007;Flanagan & Nissenbaum, 2007c). ...

Reference:

D2.1-Epic-We

... The inferential reality of computer vision AI models and real-time DSI produce privacy vulnerabilities for groups in public spaces. For the purposes of our work, Nissenbaum's theory of Privacy as Contextual Integrity (CI) helps distinguish legitimate from illegitimate information flows according to contextual norms for such groups [27]. Drawing on social theory, social philosophy, and law, CI conceives of social life as comprising distinct social domains or contexts, such as commerce, education, finance, healthcare, civic life, family, and friends [80]. ...

Reference:

Privacy of Groups in Dense Street Imagery

... 67 Risks associated with third-party cookies diverge from those generated by first-party cookies, which are minimal in comparison. 68 Notwithstanding, their use is not prohibited nor does the GDPR (nor any interpretation of the rule conducted by a DPA) mandate their absolute rolling out from the web. In any case, Google's Privacy Sandbox intended to eliminate them in favour of replicable tools which would deliver the same results in terms of advertising capabilities available for third parties. ...

Reference:

Ultra Vires Compliance as a GDPR Harm

... We also foresee risks associated with third-parties attempting to train GUMs, without consent, on a user's activity within their application. In these instances, obfuscation systems can generate interaction data to obscure genuine user behavior [51]. ...

Reference:

Creating General User Models from Computer Use

... However, the firm is not the only stakeholder whose interests matter. The decision subject's welfare is also important as part of the broader set of societal considerations [32], and indeed protecting the interests of disadvantaged consumers is the underlying motivation for fair lending interventions and regulations, including those that use the LDA. ...

Reference:

Fundamental Limits in the Search for Less Discriminatory Algorithms -- and How to Avoid Them

... For the present paper, I understand design heuristics in the sense of the term as used by Flanagan and Nissenbaum [20], who are concerned with showing alternatives to existing game design practices in order to support activist games. They propose specific principles as "design aspirations" [20] to guide the concrete development of an artifact. ...

Reference:

Design Approaches for Interactive Digital Narrative

... Optimization, often carried out as a descriptive process, underpins the creation of many computational tools like machine learning models. However, it also leads to unintended outcomes that inherently carry normative inquiries [5] that bring attention to the impact of actions on involved parties, in particular, within complex and emergent socioeconomic systems. ...

Reference:

Rethinking Optimization: A Systems-Based Approach to Social Externalities

... A central motivation behind this work is trust, which is a foundational requirement in cryptographic systems [25]. Trust is especially important when we are dealing with sensitive data such as healthcare data or financial records [1]. ...

Reference:

Extensible Post Quantum Cryptography Based Authentication

... According to this theory, privacy is defined as an appropriate flow of information according to context-dependent norms. The theory has been used to evaluate information flow in various contexts such as Facebook privacy policy [20], big data research [21], smart home [22] and contact-tracing applications [23]. ...

Reference:

Open research data and privacy violations