Monitor Prawniczy

no. 11/2024

Transparency and explainability of artificial intelligence and their functions in the GDPR and the AI Act

Dominik Lubasz

Autor jest radcą prawnym, wspólnikiem zarządzającym w Lubasz i Wspólnicy - Kancelaria Radców Prawnych, ekspertem Europejskiej Rady Ochrony Danych, członkiem Rady Naukowej Centrum Ochrony Danych Osobowych Uniwersytetu Łódzkiego, SABI - Stowarzyszenia IOD, Compliance Institute oraz członkiem komisji rewizyjnej Stowarzyszenia Prawa Nowoczesnych Technologii, ekspertem w Grupie Roboczej ds. Sztucznej Inteligencji przy Ministrze Cyfryzacji (KPRM) oraz członkiem grupy roboczej do spraw sztucznej inteligencji European Association of Data Protection Professionals

Abstract

The article provides a detailed analysis of the transparency and explainability requirements for artificial intelligence systems within the framework of the GDPR and the AI Act, emphasizing their crucial role in protecting the rights of data subjects. Transparency in AI systems, understood as providing information that allows for understanding the reasons underlying automated decisions, is a key element in fostering trust in these technologies. The article outlines the challenges associated with AI model explainability, its functions, and technical solutions that enable the system’s decisions to be explained without revealing primary causality, which is particularly important for highly complex models. The analysis highlights the role of such explainability substitutes as local output approximation and parameter-based explanations, which allow users to understand the decision-making mechanisms of AI without requiring full interpretation of a model’s internal structure.

Keywords

AI Act, personal data protection, GDPR, transparency, explainability, black box, Artificial Intelligence Act

Literature

O. Biran, C.V. Cotton, Explanation and justification in machine learning: a survey [w:] IJCAI-17 Workshop on Explainable AI (XAI), vol. 8, Nr 1/2017, s. 8-13; T. Miller, Explanation in artificial intelligence: insights from the social sciences, Artif. Intell 2019, vol. 267, s. 1-38; P. Hacker, J.-H. Passoth, Varieties of AI Explanations Under the Law. From the GDPR to the AI Act and Beyond, XXAI Beyond Explainable AI, Lecture Notes in Artificial Intelligence, Austria 2022, https://ssrn.com/abstract=3911324, s. 344; Z.C. Lipton, The mythos of model interpretability: in machine learning, the concept of interpretability is both important and slippery, Queue 2018 Nr 16(3), s. 31-57; C. Rudin, Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead, Nat. Mach. Intell. 2019, Nr 1(5), s. 206-215; A. Holzinger, R. Goebel, R. Fong, T. Moon, K.-R. Mueller, W. Same, Explainable AI Methods - A Brief Overview [w:] XXAI Beyond Explainable AI, red. A. Holzinger, R. Goebel, R. Fong, T. Moon, K.-R. Mueller, W. Samek, Lecture Notes in Artificial Intelligence, Austria 2022, s. 14; A. Holzinger, R. Goebel, R. Fong, T. Moon, K.-R. Mueller, W. Samek, Editorial [w:] XXAI Beyond Explainable AI, red. A. Holzinger, R. Goebel, R. Fong, T. Moon, K.-R. Mueller, W. Samek, Lecture Notes in Artificial Intelligence, Austria 2022, s. 4; B.P. Paal, Artificial Intelligence as a Challenge for Data Protection Law And Vice Versa [w:] The Cambridge Handbook of Responsible Artificial Intelligence, red. S. Voeneky, Ph. Kellmeyer, O. Mueller, W. Burgard, Cambridge 2022, s. 292-293; A. Holzinger, R. Goebel, R. Fong, T. Moon, K.-R. Mueller, W. Samek, Explainable AI Methods - A Brief Overview [w:] XXAI Beyond Explainable AI, red. A. Holzinger, R. Goebel, R. Fong, T. Moon, K.-R. Mueller, W. Samek, Lecture Notes in Artificial Intelligence, Austria 2022, s. 13-14; A. Bibal, M. Lognoul, A. de Streel, B. Frenay, Legal requirements on explainability in machine learning, „Artificial Intelligence Law” 2020, Nr 29(2), s. 149-169; S. Wachter, B. Mittelstadt, L. Floridi, Why a right to explanation of automated decisionmaking does not exist in the General Data Protection Regulation, SSRN Electronic Journal 2016, https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2903469; B.P. Paal [w:] Rechtshanduch Artificial Intelligence und Machine Learning, red. M. Kaulartz, T. Braegelmann, München 2020, s. 430; B. Casey, A. Farbangi, R. Vogl, Rethinking explainable machines: the GDPR’s ‘right to explanation’ debate and the rise of algorithmic audits in enterprise, „Berkeley Technology Law Journal” Nr 34/2019, s. 143; M.E. Kaminski, The right to explanation, explained, „Berkeley Technology Law Journal” Nr 34/2019, s. 189; B.P. Paal [w:] Rechtshanduch Artificial Intelligence und Machine Learning, red. M. Kaulartz, T. Braegelmann, München 2020, s. 431; T. Wischmeyer, Artificial Intelligence and Transparency. Opening the Black Box [w:] Regulating Artificial Intelligence, red. T. Wischmeyer, T. Rademacher, Cham 2020, s. 75 i n.; B.P. Paal, M. Hennemann [w:] Datenschutz-Grundverordnung. Kommentar, red. B.P. Paal, B. Pauly, Monachium 2021, kom. do art. 13; G. Mazzini [w:] Digital Revolution - New Challenges for Law, red. A. De Francesci, R. Schulze, München 2019; D. Miąsik, Zasada pierwszeństwa prawa wspólnotowego przed prawem krajowym, EPS Nr 1/2005, s. 58 i n.; P. Pałka, Ciało obce: zasady RODO a gospodarka rynkowa, „Internetowy Kwartalnik Antymonopolowy i Regulacyjny” Nr 5(11)/2022, https://ikar.wz.uw.edu.pl/images/numery/ikar_5_11/iKAR_511-22_7Palka.pdf, s. 62-84; S. Biernat, Źródła prawa Unii Europejskiej [w:] Prawo Unii Europejskiej, red. J. Barcz, Warszawa 2004, s. 256-266; T. Krügel, J. Pfeiffenbring [w:] Künstliche Intelligenz und Robotik, red. M. Ebers, Ch. Heinze, T. Krügel, B. Steinroeter, München 2020, s. 420; M. Veale, F. Zuiderveen Borgesius, Demystifying the Draft EU Artificial Intelligence Act, „Computer Law Review International” Nr 22(4)/2021, SSRN: https://ssrn.com/abstract=3896852, s. 106-108; P. Vogel, Künstliche Intelligenz und Datenschutz, Niemcy 2022, s. 107; Sh. Zuboff, Surveillance Capitalism or Democracy? The Death Match of Institutional Orders and the Politics of Knowledge in Our Information Civilization, „Organization Theory” Nr 3(3)/2022.