New paper formalizes privacy with multi-modal logic for AI

A team of researchers—Réka Markovich, Truls Pedersen, and Marija Slavkovik—published a paper titled "Understanding Privacy by Formalizing It" at the 17th International Conference on Juris-informatics (JURISIN 2023). The paper addresses a fundamental problem in AI ethics: while there is broad societal consensus on protecting privacy, the concept itself remains ambiguous and subject to multiple interpretations. To make privacy enforceable at the algorithmic level, the authors argue that we first need a precise, formal specification of what privacy means and what rights it entails.

The researchers employ multi-modal logic to formalize different theories of privacy, focusing on the right to privacy as an epistemic right within the theory of normative positions. This approach allows them to model not only what information is private but also the obligations, permissions, and prohibitions that follow from that privacy. By providing a logical framework, the paper lays groundwork for translating privacy laws and ethical principles into code—crucial for AI systems that must automatically respect user privacy, handle consent, or comply with regulations like GDPR. The work appears on arXiv (arXiv:2606.20609) and links to related tools on arXivLabs.