Focus: Responsibility and Artificial Intelligence
2026 (48) Issue 1
Editorial 
Editorial
As is evident everywhere, research into artificial intelligence has advanced by leaps and bounds in recent years, leading to a rapid increase in its use across almost all areas of life. As early as 2018, the German government of the time set out a ‘Master Plan for Artificial Intelligence’ in its coalition agreement, and the current government plans, as announced in a recently published strategy paper, to quadruple Germany’s AI capacity by 2030. Artificial intelligence is regarded as a key technology that is essential for Germany’s economic competitiveness. The development and use of AI, as well as its societal implications, are therefore a central focus of political decision-making and of debates in media and academia.
The four contributions gathered here emerged from discussions within a research network on ‘AI and Responsibility’ led by Eva Buddeberg and Fruzsina Molnár-Gábor, which was funded by the German Research Foundation (DFG) from 2022 to 2025.1 Its focus was on the question of how responsibility can be appropriately described, justified and normatively assessed in the context of artificial intelligence (AI), and under what conditions responsibility can be meaningfully attributed in complex contexts of action characterised by algorithmic systems.
Whilst AI systems are now capable of recognising complex patterns and thus independently preparing decisions or automatically controlling processes, normative assessment and the attribution of responsibility remain the preserve of human actors. This tension between technological capability and human responsibility formed the starting point for this interdisciplinary project, in which both fundamental and application-oriented aspects of responsibility were systematically examined in the light of these technological developments. Through an exchange of perspectives from philosophy, law, sociology and theology, as well as the engineering sciences, the aim – particularly from a legal perspective – was to establish a sound framework for allocating areas of responsibility when using artificial intelligence in complex technical systems. To this end, normative criteria were first identified that allow for the appropriate assessment of AI-supported decision-making processes and the definition of duties of care for the application of AI, in order to derive a liability framework that would also apply to future, unforeseeable developments in connection with AI.
We believe that the insights gained are relevant not only to the individual disciplines represented here, but also to the broader public discourse and political decision-making processes. They can provide a basis for future regulatory frameworks in the field of AI and help to create responsible and legally sound conditions for the use of algorithmic systems. Furthermore, they open new fields of research, such as the institutional embedding of prospective responsibility, the further development of liability models for autonomous systems, or the empirical investigation of societal acceptance of algorithmic decisions.
Beyond questions regarding the legal framework of responsibility and the development of specific duties of care, the focus from a philosophical and sociopolitical perspective was also, above all, on alternative conceptions of shared responsibility and issues of justice in the development and deployment of AI systems. Thus, the four contributions gathered here propose – from legal, philosophical and sociological perspectives – a more comprehensive understanding of shared or distributed joint responsibility and examine the normative and political issues arising from the use of AI systems, for example in computer-assisted selection processes.
In her article ‘Shared responsibility for the development and use of artificial intelligence’, Eva Buddeberg argues from a philosophical perspective that, going beyond liability or duty of care, we bear a discursive shared responsibility to scrutinise the development and use of AI systems against the yardstick of justice, with the aim of uncovering and counteracting injustices. She demonstrates that the now widespread use of AI in almost all areas of life tends to exacerbate existing injustices. These largely rapid developments also require us to further question, modify or, where necessary, redesign the normative framework and the associated understanding of ourselves and the world.
Drawing on insights from sociology and the studies of science and technology, Cordula Kropp and Tobias Renner develop, in their article ‘Automation, Co-Agency, and Distributed Responsibility: Caring for Hybrid Therapeutic Networks’ a concept of ‘co-agency’ as an emergent and relational property, with the aim of capturing the responsible interaction between humans and automated systems. Using a concrete example from the medical/therapeutic context, they argue that responsibility should not be attributed to individual actors – whether human or digital – but rather understood as something that arises within relational processes and should be conceived as distributed responsibility within hybrid human-machine networks. Drawing on this insight, the authors derive guidelines for a relational duty of care regarding the stability and accountability of such infrastructures.
In his article ‘Framing computational fairness and non-discrimination’, Wilfried Hinsch outlines a framework for computational fairness in the analysis of data-driven statistical discrimination. He claims that the first step is to distinguish between issues of procedural justice and issues of distributive justice. Computational procedural fairness, he argues, can only be measured by the parity of predictive values in all demographic groups. Thus, a computer-based procedure can only be non-discriminatory on the condition that it achieves the same predictive values across all demographic groups.
In ‘Perfect lawmaking and perfect legal compliance: two false ideals of normativity in governance by AI’, Klaus Günther finally analyses, from a legal-theoretical perspective, the notion that AI could enable a near-perfect normative order through, for example, data-driven lawmaking or the automated enforcement of rules. As Günther demonstrates, this vision is based on two problematic ideals: the ideal of perfectly rational lawmaking and that of complete compliance with rules. As Günther criticises, this overlooks the central importance of conflict, interpretation and democratic negotiation processes in law. This is problematic both normatively and politically, as it contributes to a technocratic conception of social governance.
As is readily apparent, the issue of ‘AI and responsibility’ is a vast field. The perspectives presented here can and should only offer an insight into what is at stake and how the problems and challenges outlined might best be addressed. Much remains open to debate, and some points may well need to be revised in the near future, not least because technological developments are transforming our entire way of life at a dramatic pace. Yet we do not, in our view, face this situation simply as passive observers. Rather, interdisciplinary and socio-political dialogue could and should make it possible – particularly through political and legal regulation – to at least contain the inherent risks and, above all, the structural injustices. In this way, the positive potential associated with technological developments could then indeed serve to better address the current problems facing our society.
Eva Buddeberg* and Fruzsina Molnár-Gábor
Table of Contents
Focus: Responsibility and Artificial Intelligence
Title: Shared Responsibility for the Development and Use of Artificial Intelligence
Author: Eva Buddeberg
Page: 5-25
Philosophical literature generally highlights three different aspects or dimensions of responsibility: 1. the attribution of authorship of actions and the liability of the actor for these actions; 2. the attribution of a duty of care for certain tasks or areas of responsibility; and finally 3. the obligation to justify one’s own behaviour and actions with good reasons. The discussion about the development and the use of artificial intelligence currently focuses on the first and, to some extent, the third aspect. However, the rapid pace of development and the now omnipresent use of AI also require ongoing critical examination and reflection on the normative framework. Responsibility must therefore be understood more broadly: with regard to the development and use of artificial intelligence, we bear a discursive shared responsibility that involves examining the development and use of AI systems against the yardstick of justice, as well as the normative framework itself that guides us in the development and use of AI systems.
Title: Framing Computational Fairness and Non-Discrimination
Author: Wilfried Hinsch
Page: 27-48
This article outlines a new framework of computational fairness for the analysis of data-driven statistical discrimination. It begins by identifying the limitations of the still dominant paradigm of fairness through unawareness. Building on this critique, the article proposes an approach that distinguishes concerns of procedural fairness from questions of distributive justice when evaluating poten tially discriminatory procedures. In contrast to a widespread view, it argues that procedural computational fairness rests on a single formal criterion. A computational procedure is non-discriminatory, it is claimed, if it achieves equal predictive values across demographic groups. In addition, two non-procedural requirements of distributive justice are advanced for selective procedures: the proportionality of individual burdens and their compatibility with a just overall social distribution of benefits and burdens.
Title: Automation, Co-Agency, and Distributed Responsibility: Caring for Hybrid Therapeutic Networks
Author: Cordula Kropp and Tobias Renner
Page: 49-70
Drawing on insights from sociology and science and technology studies, we develop an account of agency as an emergent and relational property constituted within hybrid configurations of humans and machines. We speak of co-agency to capture the ways in which responsible action is jointly produced within human-machine entanglements, with both parties demonstrating purposively intended as well as determined forms of action. By emphasizing the interdependence of sociotechnical systems, therapeutic practices, and institutional automation infrastructures, we argue that responsibility cannot be attributed to discrete actors – whether human or digital – but must instead be understood as generated within the relational processes through which co-agency emerges. This reconceptualization shifts attention from locating isolated bearers of responsibility to examining the socio-technical arrangements that structure action, thereby suggesting recommendations for ethical guidelines aimed at supporting responsible automation in healthcare.
Title: Perfect Lawmaking and Perfect Legal Compliance – Two False Ideals of Normativity in Governance by AI?
Author: Klaus Günther
Page: 71-86
Given the rapid advances in AI in predicting and controlling human behavior, the question arises as to whether the responsible use of AI might not best be regulated by law with the aid of AI itself. This question is guided by the expectation that two normative problems associated with legal norms can be definitively resolved with the help of AI: (1) that it might be possible to perfect legal norms through AI to such an extent that they provide a single correct answer for every application, and (2) that they can control the behavior of those subject to them in such a way that there is no longer any deviant behavior. This article situates these two expectations within the historical context of efforts to perfect legal norms and critically discusses the possible consequences for the democratic constitutional state.
General Part
Title: The Debate About Taxing the Wealthy
Author: Wolfgang Detel
Page: 87-110
In light of major societal challenges, the political debate on various forms of wealth taxation, or the ‘Tax the Rich’ strategy, has recently gained momentum in Germany. This article examines the arguments for and against taxing the wealthy, drawing on statistical data and some of the most significant scientific studies on current economic inequality and its history. In particular, it looks at the kinds of reasons used in the current debate. It is argued that the arguments in favor of specific forms of wealth tax outweigh the reservations.
Title: The Late and Tragic Triumph of an Economic Myth
Author: Arnis Vilks
Page: 111-131
The paper argues that the Classical Liberal Paradigm with its core tenet that the ‘invisible hand of the market’ allocates resources in a socially optimal way, has been the dominant paradigm of mainstream economics for more than 200 years, and has given rise to research associated with the names of Cournot, Marshall, Walras, Pareto, Arrow, Debreu and many others. However, the insights produced by economic theory by around 1970 must be regarded as a refutation of the invisible hand tenet. Nevertheless, propagandists of the neoliberal version of the Classical Liberal Paradigm succeeded in convincing political authorities around the world, ironically at about the same time, to implement policies suggested by the outdated paradigm. The tragedy of its late triumph lies in the fact that it has contributed to a political neglect of environmental pollution and ensuing global warming on the one hand, and to a bizarre degree of concentration of wealth and the power that comes with it, on the other hand.
Title: From the Description of ‘Wokeness’ to its Critique. Methodological Remarks on Musa Al-Gharbi’s We Have Never Been Woke
Author: Giovanni Battista Soda
Page: 133-156
This article critically examines the methodology of Musa Al-Gharbi’s We Have Never Been Woke, which offers an ambitious sociological account of ‘wokeness’ as a performative ideology deployed by elite symbolic capitalists to consolidate their privilege under the guise of progressive politics. I argue that Al-Gharbi’s account is traversed by an unresolved tension between his methodological commitments that ultimately constrains the critical reach of his analysis. Drawing on both classical and contemporary authors of Frankfurt School Critical Theory, particularly Horkheimer, Adorno, Jaeggi, and Celikates, I show that Al-Gharbi’s exclusive reliance on material-economic explanations leaves the normative and transformative capacity of ‘woke’ concepts undertheorized. Subsequently, I argue that Al-Gharbi’s descriptive account of ‘wokeness’ can be framed as a description of first order ‘woke’ practices that subsequently calls for reconstruction through a second-order self-critique. Relying on Celikates, I conclude that doing so rehabilitates the emancipatory potential of ‘woke’ concepts while exposing their conservative social entanglement, grounding a discourse that moves beyond mere demystification toward a genuinely critical account of ‘wokeness’.
Discussion
Title: Peace in Gaza?
Author: Quassim Cassam
Page: 157-177
This paper identifies five normal pre-requisites for ending a war by a formal peace agreement and argues that none of the five conditions is satisfied by both sides in the war between Israel and Hamas. Each side must be a valid interlocutor (‘interlocuteur valable’) and must recognize the other side as a valid interlocutor. There must be a degree of mutual trust and a willingness to compromise on both sides. Finally, there must be mechanisms for enforcing the agreement. Hamas’s conduct on 7 October 2023 placed it beyond the pale of dialogue and negotiation. Governments should be willing to negotiate with some terrorists but not others. Hamas in its current form belongs in the latter category. An alternative to negotiating with Hamas is to crush it with massive military force. Although some terrorist groups have been defeated in this way, Hamas is unlikely to be one of them. A more realistic objective is to continue to degrade it with a view to rendering it politically and militarily impotent. It remains to be seen whether there is any level of military pressure that would induce it to change its approach and compromise for the sake of peace.
Title: Peace in Gaza: A Response to Cassam
Author: Jeff McMahan
Page: 179-194
This article responds to Quassim Cassam’s arguments for the claim that the only viable strategy for Israel in its conflict with Hamas is to continue the war begun in October 2023 until Hamas is “crushed.” It critically examines his reasons for claiming that Hamas fails to satisfy the five conditions he states for potentially successful peace negotiations. While it may be true that Hamas fails to satisfy the conditions, it is also true that Israel fails even more egregiously to satisfy them. And even if the satisfaction of the conditions by both sides were necessary for successful negotiations, which is not the case, peace negotiations are not the best alternative to continued war. The best strategy for achieving a lasting peace is for both parties to address what Cassam calls the “root causes” of the conflict. Israel could successfully do this entirely unilaterally. And the Palestinians could also act unilaterally in a way that would be the best they could do to eliminate the fundamental source of the conflict.
Title: The Meaning of Social Ownership, the Priority of Cooperativism, and the Unpleasant Need for Private Shareholding: A Reply to Claassen
Author: Hannes Kuch
Page: 195-205
This paper defends the model of Plural Cooperativism against criticism. Plural Cooperativism is a society-wide economic system of cooperative ownership that allows for limited private shareholding, counterbalanced by public shareholding and a more democratic reallocation of control rights, while universal inheritances enable broad-based investment in cooperatives. In response to Rutger Claassen’s critique, the paper clarifies the normative foundations of social ownership and argues that cooperatives can serve a wide range of values beyond workplace democracy, especially when a public shareholder complements them. It further defends the focus on cooperatives against Claassen’s proposal to rely more strongly on trust-owned firms, since the latter exhibit a structural deficit in workplace democracy and impose overly demanding altruistic requirements. Finally, it justifies the opportunity for private shareholding by addressing issues of entry costs, financing, and portfolio diversification, and it argues for a strong role of the public shareholder by showing that disinvestment threats from private shareholders persist under Plural Cooperativism.