Limits of Optimisation - Structures, Conditions and Limits of Algorithmic Legal Implementation in a Democratic Society

• Applicant: Professor Dr. Timo Rademacher
• Subject Area: Public Law
• Term: since 2025
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 559134250

To this day, it is an undisputed fact: The realisation of law is contingent. It depends on so many variables that the adherence to and enforcement of the law often appears to happen more or less randomly (Lübbe-Wolff). However, the technologies currently referred to as "artificial intelligence" are likely to change this fact: The realisation of law will, most likely, become increasingly optimised through technological means. In light of these developments, I pursue three objectives in this study, as outlined in the subtitle: In Chapter 1, I explore the AI-based possibilities/structures that exist or are likely to exist soon to optimise the realisation of law. The possibilities range from assistance systems that facilitate compliance with the law (Compliance Assistance Technologies) and optimised surveillance technologies (Intelligent Surveillance) to structures that make unlawful behaviour impossible (Impossibility Structures). An in-between structure that has so far been completely neglected in academic discussions is what I call Justification Structures, which link an assistance system with a surveillance system, meaning they allow a (prima facie) breach of law, but only in exchange for a corresponding report to authorities or other rights holders. Examples of all these categories already exist, and given the potential of AI future possibilities appear – from a factual point of view – virtually limitless. Thus, the legal limits are becoming increasingly important. Yet before I address the legal limits of technologies in detail in Chapter 3, I first examine, in Chapter 2, the conditions under which the use of AI for the realisation of law or its optimisation may be permissible. Here, several questions are to be addressed: Can technology interpret/apply law at all? Which standard should apply in this context – that of a professional legal practitioner or of a legal layperson? Can/must/should legal realisation technologies alone be capable of interpreting the law "correctly", or is it sufficient if they do so in hybrid architectures of human-machine cooperation? Once these conditions have been clarified, Chapter 3 can turn to the core question, i. e. the question of limits: If it is possible to realise law through technology with precision, producing as few false positives or false negatives as possible, is there still any reason to refrain from such technological optimisation? To put it differently, or more provocatively: Does breaching the law have an intrinsic value that should be preserved? Here, I examine various doctrinal foundations within the German constitution, considering the relevant impulses from legal theory, legal philosophy, legal economics, and legal sociology. Ultimately, I argue for the intrinsic value of legal breaches, which, based on my analysis, can be grounded in the principle of democracy. I also consider whether and to what extent randomization (as suggested e. g. by Christoph Möllers, among others) can help to realise/preserve this value.

• DFG Programme: Publication Grants