KNOW.THY.AI
Access Evidence, Uncover Causality - Informational Transparency as Cornerstone of A Needs-Based Theory of Procedural Fairness in AI Liability Disputes
KNOW.THY.AI
Access Evidence, Uncover Causality - Informational Transparency as Cornerstone of A Needs-Based Theory of Procedural Fairness in AI Liability Disputes
Seeking to prevent harm caused by Artificial Intelligence (AI) systems, national and supranational regulators chose a priori technical standardization (in the European Union - EU, the AI Act, COM (2021) 206 final). In doing so, they paid little attention to the procedural safeguards that human agents (programmers, deployers) should benefit from when AI-related harm materializes.
Though procedural AI regulation has recently begun to emerge (in the EU, AI Liability Directive, COM(2022) 496 final), neither scholarship nor the existing court practice provide definitive solutions on the types of evidence that litigants should be legally entitled to access, adduce and explain, when they discuss fault and causation in AI liability cases. The Know.Thy.AI project (48 months, 3 Work Packages - WP) will fill this gap. First (WP1), it will induce the most relevant and probative evidence in the field of AI liability based on the consultation of at least 140 expert respondents (legal practitioners, courts, AI programmers) selected from two European organizations (EU, Council of Europe) as well as five European countries (Belgium, France, Netherlands, Germany and Italy). By determining their probative values using the Analysis of variance (ANOVA method), KNOW.THY.AI will create a taxonomy of evidentiary items which will, second, serve as basis for: 1. regulatory proposals of a framework of procedural abilities that litigants must be entitled to so that AI liability disputes may be effectively adjudicated (WP2), 2. the creation of an open access, state-of-the art Natural-Language Processing (NLP) AI, which could be used to audit the level of informational transparency in future causal scenarios involving AI systems (WP3).
Know.Thy.AI will suggest a model for a needs-based theory of procedural fairness in the field of AI liability, by highlighting the procedural channels that should be made available so that judicial redress in AI liability cases can be effective.
Principal investigator (PI): Ljupcho GROZDANOVSKI
Co-supervisor: Julien CABAY
Ph.D candidate: TBA
Ph.D candidate: TBA
