Łukasz Górski
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Research Interests
Łukasz Górski has a total of 48 co-authors. Here are the top co-authors by number of shared publications:
- Jedrzej M. Nowosielski
Interdisciplinary Centre for Mathematical and Computational Modelling, University of Warsaw, Warsaw, Poland
4 shared publications - Antoni Moszyński
Interdisciplinary Centre for Mathematical and Computational Modelling, University of Warsaw
3 shared publications - Franciszek Rakowski
Interdisciplinary Centre for Mathematical and Computational Modelling, University of Warsaw
3 shared publications - Karol Niedzielewski
Interdisciplinary Centre for Mathematical and Computational Modelling (ICM) University of Warsaw Poland
3 shared publications - Maciej Radwan
Interdisciplinary Centre for Mathematical and Computational Modelling, University of Warsaw
3 shared publications - Magdalena Gruziel-Słomka
Interdisciplinary Centre for Mathematical and Computational Modelling, University of Warsaw
3 shared publications
Latest Publications
- Open Access
- 1 July 2026
- Published by University of Warsaw in Studia Iuridica
- Vol. 110 (110) , 107-127
AGAINST EXPLAINABLE ARTIFICIAL INTELLIGENCE IN LAW: WHY JUSTIFIABLE AI MATTERS. A CREDIT SCORING EXAMPLE
Artificial intelligence-based solutions offer new efficiency-increasing possibilities in many applications, including credit scoring. Yet, the increasing sophistication of machine-learning models in use raises concerns regarding many of their aspects, explainability notwithstanding. We review the relevant EU legal background and integrate this review with insights from technical sciences to interpret relevant legal provisions in the light of technological possibilities. We reject the narrow interpretations of the right to explanation and suggest the broad one, which encompasses not only technical explanations but also a legal justification as the only one that allows for safeguarding the creditors’ rights in an operative manner.
- 1 July 2026
- Published by University of Warsaw in Studia Iuridica
- Vol. 110 (110) , 107-127
Open AccessAGAINST EXPLAINABLE ARTIFICIAL INTELLIGENCE IN LAW: WHY JUSTIFIABLE AI MATTERS. A CREDIT SCORING EXAMPLE
Artificial intelligence-based solutions offer new efficiency-increasing possibilities in many applications, including credit scoring. Yet, the increasing sophistication of machine-learning models in use raises concerns regarding many of their aspects, explainability notwithstanding. We review the relevant EU legal background and integrate this review with insights from technical sciences to interpret relevant legal provisions in the light of technological possibilities. We reject the narrow interpretations of the right to explanation and suggest the broad one, which encompasses not only technical explanations but also a legal justification as the only one that allows for safeguarding the creditors’ rights in an operative manner.
- 19 February 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IT Professional
- Vol. 27 (1) , 34-40
Right to Explanation in Large Language Models: Lessons from the European Union AI Act and General Data Protection Regulation
Innovations in machine learning-based artificial intelligence (AI), culminating in the development of large-scale language models, became a watershed moment for AI research. AI and law research follows suit, and the new regulatory framework is being prepared. In this article, we explore the legal background for explainability of large-scale foundational models, as brought about by the European Union AI Act.
- 19 February 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IT Professional
- Vol. 27 (1) , 34-40
Right to Explanation in Large Language Models: Lessons from the European Union AI Act and General Data Protection Regulation
Innovations in machine learning-based artificial intelligence (AI), culminating in the development of large-scale language models, became a watershed moment for AI research. AI and law research follows suit, and the new regulatory framework is being prepared. In this article, we explore the legal background for explainability of large-scale foundational models, as brought about by the European Union AI Act.
- Open Access
- 7 May 2024
- Published by Springer Nature in Artificial Intelligence and Law
- Vol. 33 (3) , 551-579
Exploring explainable AI in the tax domain
This paper analyses whether current explainable AI (XAI) techniques can help to address taxpayer concerns about the use of AI in taxation. As tax authorities around the world increase their use of AI-based techniques, taxpayers are increasingly at a loss about whether and how the ensuing decisions follow the procedures required by law and respect their substantive rights. The use of XAI has been proposed as a response to this issue, but it is still an open question whether current XAI techniques are enough to meet existing legal requirements. The paper approaches this question in the context of a case study: a prototype tax fraud detector trained on an anonymized dataset of real-world cases handled by the Buenos Aires (Argentina) tax authority. The decisions produced by this detector are explained through the use of various classification methods, and the outputs of these explanation models are evaluated on their explanatory power and on their compliance with the legal obligation that tax authorities provide the rationale behind their decision-making. We conclude the paper by suggesting technical and legal approaches for designing explanation mechanisms that meet the needs of legal explanation in the tax domain.
- 7 May 2024
- Published by Springer Nature in Artificial Intelligence and Law
- Vol. 33 (3) , 551-579
Open AccessExploring explainable AI in the tax domain
This paper analyses whether current explainable AI (XAI) techniques can help to address taxpayer concerns about the use of AI in taxation. As tax authorities around the world increase their use of AI-based techniques, taxpayers are increasingly at a loss about whether and how the ensuing decisions follow the procedures required by law and respect their substantive rights. The use of XAI has been proposed as a response to this issue, but it is still an open question whether current XAI techniques are enough to meet existing legal requirements. The paper approaches this question in the context of a case study: a prototype tax fraud detector trained on an anonymized dataset of real-world cases handled by the Buenos Aires (Argentina) tax authority. The decisions produced by this detector are explained through the use of various classification methods, and the outputs of these explanation models are evaluated on their explanatory power and on their compliance with the legal obligation that tax authorities provide the rationale behind their decision-making. We conclude the paper by suggesting technical and legal approaches for designing explanation mechanisms that meet the needs of legal explanation in the tax domain.
- Open Access
- 7 December 2023
- Published by IOS Press
- p. 275-280
Challenges in Adapting LLMs for Transparency: Complying with Art. 14 EU AI Act
This paper investigates the technical potential of developing explainable transformer-based large language models (LLMs) that meet users’ needs and comply with regulations such as the EU AI Act. We implement and evaluate a proof-of-concept explainable LLM-based system. Furthermore, we demonstrate its potential for regulatory compliance and identify technical challenges to be tackled.
- 7 December 2023
- Published by IOS Press
- p. 275-280
Open AccessChallenges in Adapting LLMs for Transparency: Complying with Art. 14 EU AI Act
This paper investigates the technical potential of developing explainable transformer-based large language models (LLMs) that meet users’ needs and comply with regulations such as the EU AI Act. We implement and evaluate a proof-of-concept explainable LLM-based system. Furthermore, we demonstrate its potential for regulatory compliance and identify technical challenges to be tackled.
- Open Access
- 14 November 2023
- Published by Springer Nature in Research Square
Forecasting SARS-CoV-2 Epidemic Dynamic in Poland with the pDyn Agent-Based Model
We employ pDyn, an agent-based epidemiological model, to forecast the fourth wave of the SARS-CoV-2 epidemic, primarily driven by the Delta variant, in Polish society. The model captures spatiotemporal dynamics of the epidemic spread, predicting disease-related states based on pathogen properties and behavioral factors. We assess pDyn's validity, encompassing pathogen variant succession, immunization level, and the proportion of vaccinated among confirmed cases. We evaluate its predictive capacity for pandemic dynamics, including wave peak timing, magnitude, and duration for confirmed cases, hospitalizations, ICU admissions, and deaths, nationally and regionally in Poland. Validation involves comparing pDyn's estimates with real-world data (excluding data used for calibration) to evaluate whether pDyn accurately reproduced the epidemic dynamics up to the simulation time. To assess the accuracy of pDyn's predictions, we compared simulation results with real-world data acquired after the simulation date. The findings affirm pDyn's accuracy in forecasting and enhancing our understanding of epidemic mechanisms.
- 14 November 2023
- Published by Springer Nature in Research Square
Open AccessForecasting SARS-CoV-2 Epidemic Dynamic in Poland with the pDyn Agent-Based Model
We employ pDyn, an agent-based epidemiological model, to forecast the fourth wave of the SARS-CoV-2 epidemic, primarily driven by the Delta variant, in Polish society. The model captures spatiotemporal dynamics of the epidemic spread, predicting disease-related states based on pathogen properties and behavioral factors. We assess pDyn's validity, encompassing pathogen variant succession, immunization level, and the proportion of vaccinated among confirmed cases. We evaluate its predictive capacity for pandemic dynamics, including wave peak timing, magnitude, and duration for confirmed cases, hospitalizations, ICU admissions, and deaths, nationally and regionally in Poland. Validation involves comparing pDyn's estimates with real-world data (excluding data used for calibration) to evaluate whether pDyn accurately reproduced the epidemic dynamics up to the simulation time. To assess the accuracy of pDyn's predictions, we compared simulation results with real-world data acquired after the simulation date. The findings affirm pDyn's accuracy in forecasting and enhancing our understanding of epidemic mechanisms.
- 23 September 2022
- Published by Springer Nature
- p. 221-238
Requirements for Tax XAI Under Constitutional Principles and Human Rights
- 23 September 2022
- Published by Springer Nature
- p. 221-238
Requirements for Tax XAI Under Constitutional Principles and Human Rights
- 27 November 2021
- Published by Springer Nature
- p. 154-168
Towards Grad-CAM Based Explainability in a Legal Text Processing Pipeline. Extended Version
- 27 November 2021
- Published by Springer Nature
- p. 154-168
Towards Grad-CAM Based Explainability in a Legal Text Processing Pipeline. Extended Version
- Open Access
The efficiency of dynamic regional lockdown approach in controlling the COVID-19 epidemic. Insights from the agent-based epidemiological model for Poland
In this work properties of the dynamic regional lockdown approach to suppress the COVID-19 epidemic spread in Poland were investigated. In particular, an agent based model was used with the aim to indicate an optimal lockdown strategy, defined here as the one which minimizes mean lockdown time over regional unit provided health service is not overwhelmed. With this approach the lockdown extent was also considered by varying restrictions between complete regional school closure and/or significant social distancing in semi-public spaces. In result, a cooperative effect was discovered in the case when closure of schools was accompanied by severe restrictions of social contacts in semi-public spaces. Moreover, the regional lockdown approach implemented here on the level of counties (units of population around 100k) proofed to be successful, that is allowed to identify optimal entrance and release thresholds for lockdown. The authors believe that until significant portion of population is vaccinated such a strategy might be applied.
Open AccessThe efficiency of dynamic regional lockdown approach in controlling the COVID-19 epidemic. Insights from the agent-based epidemiological model for Poland
In this work properties of the dynamic regional lockdown approach to suppress the COVID-19 epidemic spread in Poland were investigated. In particular, an agent based model was used with the aim to indicate an optimal lockdown strategy, defined here as the one which minimizes mean lockdown time over regional unit provided health service is not overwhelmed. With this approach the lockdown extent was also considered by varying restrictions between complete regional school closure and/or significant social distancing in semi-public spaces. In result, a cooperative effect was discovered in the case when closure of schools was accompanied by severe restrictions of social contacts in semi-public spaces. Moreover, the regional lockdown approach implemented here on the level of counties (units of population around 100k) proofed to be successful, that is allowed to identify optimal entrance and release thresholds for lockdown. The authors believe that until significant portion of population is vaccinated such a strategy might be applied.
- 2 August 2021
- Published by Wiley in Concurrency and Computation: Practice and Experience
- Vol. 35 (15) , e6536
Performance evaluation of Java/PCJ implementation of parallel algorithms on the cloud (extended version)
Large‐scale computing and data processing with cloud resources is gaining popularity. However, the usage of the cloud differs from traditional high‐performance computing (HPC) systems and both algorithms and codes have to be adjusted. This work is often time‐consuming and performance is not guaranteed. To address this problem we have developed the PCJ library (parallel computing in Java), a novel tool for scalable HPC and big data processing in Java. In this article, we present a performance evaluation of parallel applications implemented in Java using the PCJ library. The performance evaluation is based on the examples of highly scalable applications of different characteristics focusing on CPU, communication or I/O. They run on the traditional HPC system and Amazon web services Cloud as well as Linaro Developer Cloud. For the clouds, we have used Intel x86 and ARM processors for running Java codes without changing any line of the program code and without the need for time‐consuming recompilation. Presented applications have been parallelized using the partitioned global address space programming model and its realization in the PCJ library. Our results prove that the PCJ library, due to its performance and ability to create simple portable code, has great promise to be successful for the parallelization of various applications and run them on the cloud with a performance close to HPC systems.
- 2 August 2021
- Published by Wiley in Concurrency and Computation: Practice and Experience
- Vol. 35 (15) , e6536
Performance evaluation of Java/PCJ implementation of parallel algorithms on the cloud (extended version)
Large‐scale computing and data processing with cloud resources is gaining popularity. However, the usage of the cloud differs from traditional high‐performance computing (HPC) systems and both algorithms and codes have to be adjusted. This work is often time‐consuming and performance is not guaranteed. To address this problem we have developed the PCJ library (parallel computing in Java), a novel tool for scalable HPC and big data processing in Java. In this article, we present a performance evaluation of parallel applications implemented in Java using the PCJ library. The performance evaluation is based on the examples of highly scalable applications of different characteristics focusing on CPU, communication or I/O. They run on the traditional HPC system and Amazon web services Cloud as well as Linaro Developer Cloud. For the clouds, we have used Intel x86 and ARM processors for running Java codes without changing any line of the program code and without the need for time‐consuming recompilation. Presented applications have been parallelized using the partitioned global address space programming model and its realization in the PCJ library. Our results prove that the PCJ library, due to its performance and ability to create simple portable code, has great promise to be successful for the parallelization of various applications and run them on the cloud with a performance close to HPC systems.