Enrico Panai
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
Enrico Panai has a total of 25 co-authors. Here are the top co-authors by number of shared publications:
- Aleksandra Łapko
Maritime University of Szczecin
3 shared publications - Roma Strulak-Wójcikiewicz
Faculty of Economics and Transport Engineering, Maritime University of Szczecin, Wały Chrobrego 1/2, 70-500 Szczecin, Poland
3 shared publications Alessio Tartaro
2 shared publications- Alessio Tartaro
Department of Humanities and Social Sciences, University of Sassari, Sassari, Italy
2 shared publications - Ewa Hącia
Akademia Morska w Szczecinie
2 shared publications - Kevser Çınar
Necmettin Erbakan University
2 shared publications
Latest Publications
- Open Access
- 26 June 2026
- Published by Elsevier in SSRN Electronic Journal
Decision Problems for Regulatees under the EU AI Act: Contested Values, Uncertain Evidence, and the Limits of Standardisation
Based on the New Legislative Framework (NLF) approach, the AI Act relies on harmonised standards to provide technical specifications for the implementation of its essential requirements for high-risk AI systems. This paper argues that these requirements pose fundamental "decision problems" for regulatees, requiring value judgments and evidence assessment under uncertainty for their implementation. Unlike mature NLF fields with established methodologies and value consensus, the AI domain is characterised by contested values and uncertain evidence, significantly limiting the ability of traditional standardisation to provide clear, universally applicable solutions. This creates uncertainty for regulatees, complicates conformity assessment and enforcement, and risks undermining the overall regulatory effectiveness of the AI Act. In response to these challenges, this paper proposes supplementing standardisation with a procedural approach, such as documented "AI Act Compliance Cases," to compel transparent articulation and justification of regulatees' decisions. This enhances auditability and manageability, bolstering the Act's capacity to achieve its health, safety and fundamental rights objectives despite inherent complexities.
- 26 June 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessDecision Problems for Regulatees under the EU AI Act: Contested Values, Uncertain Evidence, and the Limits of Standardisation
Based on the New Legislative Framework (NLF) approach, the AI Act relies on harmonised standards to provide technical specifications for the implementation of its essential requirements for high-risk AI systems. This paper argues that these requirements pose fundamental "decision problems" for regulatees, requiring value judgments and evidence assessment under uncertainty for their implementation. Unlike mature NLF fields with established methodologies and value consensus, the AI domain is characterised by contested values and uncertain evidence, significantly limiting the ability of traditional standardisation to provide clear, universally applicable solutions. This creates uncertainty for regulatees, complicates conformity assessment and enforcement, and risks undermining the overall regulatory effectiveness of the AI Act. In response to these challenges, this paper proposes supplementing standardisation with a procedural approach, such as documented "AI Act Compliance Cases," to compel transparent articulation and justification of regulatees' decisions. This enhances auditability and manageability, bolstering the Act's capacity to achieve its health, safety and fundamental rights objectives despite inherent complexities.
- Open Access
- 23 March 2026
- Published by Springer Nature in AI and Ethics
- Vol. 6 (2)
The ghost in the machine speaks with an American accent: cultural value drift in early GPT-3 and the case for pluralist evaluation of generative AI
Early large language models (LLMs) were released with minimal alignment, offering a rare view of how generative systems reframe the ethical values embedded in human texts. We examine outputs from a 2021 version of OpenAI’s base GPT-3, prompting it to summarise culturally diverse source materials (laws, political speeches, and philosophical works) and interpreting results through a descriptive, moral value pluralist lens. Where possible, we contextualise outputs with cross-national datasets such as the World Values Survey. We document recurring value drift: Australia’s firearm policy is recast as a threat to liberty; de Beauvoir’s feminist critique becomes gender-essentialist dating advice; and Merkel’s humanitarian appeal is recast as immigration control. In contrast, multilateral documents (UN/UNESCO) exhibit greater value stability, suggesting consensus-crafted language can buffer against cultural mutation. We argue that these early behaviours (observed before extensive fine-tuning and safety layers) provide a baseline for understanding how training distributions shape normative framing. Our contribution is twofold: (1) empirical evidence that value drift can invert or overwrite encoded values along predictable cultural axes, and (2) a pluralist, descriptive evaluation method that surfaces whose values dominate and when. We conclude with implications for culturally inclusive evaluation and alignment in contemporary LLMs.
- 23 March 2026
- Published by Springer Nature in AI and Ethics
- Vol. 6 (2)
Open AccessThe ghost in the machine speaks with an American accent: cultural value drift in early GPT-3 and the case for pluralist evaluation of generative AI
Early large language models (LLMs) were released with minimal alignment, offering a rare view of how generative systems reframe the ethical values embedded in human texts. We examine outputs from a 2021 version of OpenAI’s base GPT-3, prompting it to summarise culturally diverse source materials (laws, political speeches, and philosophical works) and interpreting results through a descriptive, moral value pluralist lens. Where possible, we contextualise outputs with cross-national datasets such as the World Values Survey. We document recurring value drift: Australia’s firearm policy is recast as a threat to liberty; de Beauvoir’s feminist critique becomes gender-essentialist dating advice; and Merkel’s humanitarian appeal is recast as immigration control. In contrast, multilateral documents (UN/UNESCO) exhibit greater value stability, suggesting consensus-crafted language can buffer against cultural mutation. We argue that these early behaviours (observed before extensive fine-tuning and safety layers) provide a baseline for understanding how training distributions shape normative framing. Our contribution is twofold: (1) empirical evidence that value drift can invert or overwrite encoded values along predictable cultural axes, and (2) a pluralist, descriptive evaluation method that surfaces whose values dominate and when. We conclude with implications for culturally inclusive evaluation and alignment in contemporary LLMs.
- Open Access
- 1 January 2025
- Published by Elsevier in SSRN Electronic Journal
Leveraging Technical Standards within AI Regulatory Sandboxes: Challenges and Opportunities
The AI Act relies on harmonized standards to provide cutting-edge technical solutions for implementing regulatory requirements, streamlining conformity assessment procedures, and offering a presumption of compliance for high-risk AI system providers. This paper examines the multifaceted role of standards within the AI Act, with a particular emphasis on their interaction with regulatory sandboxes. It begins with a comprehensive overview of the standards’ role in the regulation, followed by an in-depth analysis of the synergy between regulatory sandboxes and technical standards. The study then explores the challenges and opportunities arising from the use of harmonized standards within the sandbox environment. Finally, it presents targeted recommendations for competent authorities, European standardization organizations (ESOs), and AI providers to maximize the potential of standards in regulatory sandbox environments. This research contributes to the ongoing discourse on effective AI governance and the practical implementation of the AI Act.
- 1 January 2025
- Published by Elsevier in SSRN Electronic Journal
Open AccessLeveraging Technical Standards within AI Regulatory Sandboxes: Challenges and Opportunities
The AI Act relies on harmonized standards to provide cutting-edge technical solutions for implementing regulatory requirements, streamlining conformity assessment procedures, and offering a presumption of compliance for high-risk AI system providers. This paper examines the multifaceted role of standards within the AI Act, with a particular emphasis on their interaction with regulatory sandboxes. It begins with a comprehensive overview of the standards’ role in the regulation, followed by an in-depth analysis of the synergy between regulatory sandboxes and technical standards. The study then explores the challenges and opportunities arising from the use of harmonized standards within the sandbox environment. Finally, it presents targeted recommendations for competent authorities, European standardization organizations (ESOs), and AI providers to maximize the potential of standards in regulatory sandbox environments. This research contributes to the ongoing discourse on effective AI governance and the practical implementation of the AI Act.
- 14 December 2024
- Published by Springer Nature in Ethics and Information Technology
- Vol. 27 (1) , 1-28
Nullius in Explanans: an ethical risk assessment for explainable AI
- 14 December 2024
- Published by Springer Nature in Ethics and Information Technology
- Vol. 27 (1) , 1-28
Nullius in Explanans: an ethical risk assessment for explainable AI
- Top Cited Papers
- 22 October 2024
- Published by Springer Nature in Nature Human Behaviour
- Vol. 8 (10) , 1829-1830
The case for human–AI interaction as system 0 thinking
- 22 October 2024
- Published by Springer Nature in Nature Human Behaviour
- Vol. 8 (10) , 1829-1830
Top Cited PapersThe case for human–AI interaction as system 0 thinking
- 3 January 2024
- Published by Springer Nature in AI and Ethics
- Vol. 4 (1) , 105-112
AI risk assessment using ethical dimensions
- 3 January 2024
- Published by Springer Nature in AI and Ethics
- Vol. 4 (1) , 105-112
AI risk assessment using ethical dimensions
- 4 September 2023
- Published by Springer Nature in AI & SOCIETY
- Vol. 39 (6) , 2647-2665
The latent space of data ethics
- 4 September 2023
- Published by Springer Nature in AI & SOCIETY
- Vol. 39 (6) , 2647-2665
The latent space of data ethics
- Open Access
- 1 January 2023
- Published by Globethics Publications
- p. 97-110
Data ethics officer : a role to build
With Artificial Intelligence (AI) and the development and spread of systems that fall into this category, ethics has returned to the forefront. 149 To the extent that the noun 'ethics', the adjective 'ethical' and the adverb 'ethically' are used in most texts on the governance and policies of AI
- 1 January 2023
- Published by Globethics Publications
- p. 97-110
Open AccessData ethics officer : a role to build
With Artificial Intelligence (AI) and the development and spread of systems that fall into this category, ethics has returned to the forefront. 149 To the extent that the noun 'ethics', the adjective 'ethical' and the adverb 'ethically' are used in most texts on the governance and policies of AI
- 10 November 2022
- Published by Springer Nature in Digital Society
- Vol. 1 (3) , 1-20
The Self-Synchronisation of AI Ethical Principles
- 10 November 2022
- Published by Springer Nature in Digital Society
- Vol. 1 (3) , 1-20
The Self-Synchronisation of AI Ethical Principles
- Open Access
- 23 December 2021
- Published by MDPI AG in Sustainability
- Vol. 14 (1) , 134
Eco-Friendly Tourism Decision Making during COVID-19—Sailing Tourism Example
In 2020, tourism was highly affected by COVID-19 and its restrictions, such as tourist traffic. Decisions related to trips were made in a state of high risk and uncertainty. This article’s main aim is to present the results of research on decision making by people practising sailing tourism during COVID-19. The survey was conducted in the first quarter of 2021 on 580 sailors from Poland, Germany, Croatia, Italy, France, and Turkey. This is interesting because of the specificity of this form of nautical tourism, which is characterised by high flexibility in the planning and implementation of the cruise. Sailing tourism is also environmentally friendly due to the type of propulsion used and the low noise levels generated. Research has shown that country-specific travel restrictions impact sailing tourism and cruise decisions. The obtained results are important for developing sailing tourism and may contribute to rationalising decisions taken during crises.
- 23 December 2021
- Published by MDPI AG in Sustainability
- Vol. 14 (1) , 134
Open AccessEco-Friendly Tourism Decision Making during COVID-19—Sailing Tourism Example
In 2020, tourism was highly affected by COVID-19 and its restrictions, such as tourist traffic. Decisions related to trips were made in a state of high risk and uncertainty. This article’s main aim is to present the results of research on decision making by people practising sailing tourism during COVID-19. The survey was conducted in the first quarter of 2021 on 580 sailors from Poland, Germany, Croatia, Italy, France, and Turkey. This is interesting because of the specificity of this form of nautical tourism, which is characterised by high flexibility in the planning and implementation of the cruise. Sailing tourism is also environmentally friendly due to the type of propulsion used and the low noise levels generated. Research has shown that country-specific travel restrictions impact sailing tourism and cruise decisions. The obtained results are important for developing sailing tourism and may contribute to rationalising decisions taken during crises.