Workshop

Authorless Texts? Generative AI, Authorship, and the Future of Writing

My presentation will be about how to think about – and whether we need to rethink – the idea of authorship as many people (students, researchers, artists, novelists, etc.) start to rely more and more heavily on large language model (LLM)-based generative AI technologies when they produce texts. In the small but growing literature on this topic, the following views are all represented: (1) LLMs can be authors of texts; (2) LLMs can be co-authors of texts; (3) when people use LLMs, they are always the authors of the resulting texts; (4) when people use LLMs, they are co-authors, but not sole authors; and (5) texts whose production has relied heavily on LLMs should be considered as authorless texts. I will discuss these different views, doing so with reference to the ideas that authors are accountable for their texts, that writing is a form of agency, and that if texts are good, authors deserve credit for their texts.

Seeing the forest for the trees: responsibility beyond the ‘gap’ in autonomous warfare

The increasing deployment of artificial intelligence (AI) across domains, including the military, has intensified debates about the responsible use of autonomous systems. Central to these discussions is the question of moral responsibility: when highly autonomous systems make decisions with grave consequences, who—if anyone—can be held responsible? Some scholars argue that such systems generate a “responsibility gap,” in which traditional frameworks of moral and legal accountability appear to break down.

This presentation offers a critical re-examination of the responsibility gap debate. Rather than treating the problem as a wholly novel consequence of technological advancement, it situates contemporary discussions within the longer history of responsibility theory. By disentangling the existing literature, the talk identifies three key fault lines: between those who affirm or deny the existence of responsibility gaps; those who view such gaps as a fundamentally new moral problem versus those who do not; and those who are optimistic about closing these gaps versus those who regard them as unsolvable. Building on this analysis, the presentation argues that much of the apparent novelty of the responsibility gap dissolves when examined through the lens of longstanding philosophical questions—particularly concerning the role of reactive attitudes and the metaphysical grounding of responsibility. On this view, the challenge posed by autonomous systems lies less in the technology itself and more in tensions within our existing conceptual frameworks. Ultimately, the presentation argues that progress in the field of responsibility in the age of autonomous systems does not depend on abandoning existing theories, but on clarifying and, where necessary, revising the conceptual foundations of traditional theories of responsibility.

From Shared Control to Shared Confusion. Delusions of Agency and Responsibility Gaps in Human-AI Systems

Shared control between humans and automated intelligent systems can give rise to responsibility gaps: situations in which it is unclear whom to assign responsibility. Current debates often focus on issues of distribution: who had control, who had knowledge, and how systems can be made explainable or contestable. Recent accounts of so-called “Meaningful Human Control” have been trying to redefine form of control based on normative requirements that would at least partially amend those gaps. However, we argue that even such theories fall short of accounting for the psychological experience of agency, and its effect on assumed and attributed responsibility. In fact, the interaction between human controllers and AI-automated systems can affect a controller’s sense of agency (SoA): the experience of being in control over one’s actions and their effect on the surrounding world. We show that in shared-control contexts, a controller’s SoA can lead to unfair (excessive/insufficient) responsibility (self-)attributions by further aggravating five distinct types of responsibility gaps: culpability, moral accountability, public accountability, active responsibility and vulnerability. As a consequence, it is harder for both individuals and institutions to meaningfully assign and assume responsibility. We conclude that SoA must be studied and included in a normative set of requirements for meaningful human control and responsibility. In order to facilitate this, we outline a preliminary set of requirements, in the form of design-oriented and institutional directions for supporting calibrated agency in collaborative AI environments.

Defining AI-based Autonomy Gaps

Starting with responsibility gaps, there has been an evolving discussion around various gaps related to AI technology, concerning control, meaningfulness, retribution, achievement, and authorship. This raises the need to define AI-based gaps, more generally, and to identify other kinds of gaps, more specifically. Among the growing set of AI-based gaps, there is one case that has remained mostly neglected—namely, so-called “autonomy gaps.” Given the importance of human autonomy in AI ethics, autonomy gaps demand further scrutiny. This paper aims to remedy these shortcomings by: First, providing a definitional schema of AI-based gaps. Second, proposing a definition of autonomy gaps based on the proposed definitional schema. Third, showcasing the existence of autonomy gaps, while distinguishing them from other related impacts of AI on human autonomy. Fourth and finally, further motivating autonomy gaps’ conceptual and ethical importance for key AI ethics debates. In so doing, this paper also contributes to clarifying the challenges of AI for human autonomy, as well as the emergence of other AI-based gaps.

AI achievement gaps: how (not) to deal with them

In this talk, I will present the problem of AI achievement gaps and illustrate some of its troubling consequences (Danaher & Nyholm 2021). I will examine some problematic ways of dealing with this problem: analysing the problem away (Tigard 2021), ignoring it, or filling the gap with responsibility of AI systems (Kieval 2024, criticised in Vacek 2025a). I will then propose some promising ways forward: filling it with collective, vicarious, or even proxy responsibility of human stakeholders (the former two are defended in Vacek 2025b). 

References

 Danaher, J., & Nyholm, S. (2021). Automation, work and the achievement gap. AI and Ethics, 1(3), 227–237.

Kieval, P. H. (2024). Artificial achievements. Analysis, 84(1), 32-41.

Tigard, D. W. (2021). Workplace automation without achievement gaps: A reply to Danaher and Nyholm. AI and Ethics, 1(4), 611–617.

Vacek, D. (2025a). Against artificial achievements. Analysis, 85(3), 690–698.

Vacek, D. (2025b). Meeting the AI achievement challenge: collective and vicarious achievements. Ethics and Information Technology, 27(2), 25.