Abstracts

The Moral Implications of Epistemic Dependence on AI

Stefan Buijsman (TU Delft)

The advances in machine learning-based systems in medicine give rise to pressing epistemological and ethical questions. Clinical decisions are increasingly taken in highly digitized work environments, which we call artificial epistemic niches. By considering the case of ML systems in life-critical healthcare settings, we investigate (i) when users' reliance on these systems can be characterized as epistemic dependence and (ii) how this dependence turns into what we refer to as problematic epistemic dependence of clinical professionals. Problematic epistemic dependence occurs when the impossibility of critically assessing the soundness of a system’s output in situ implies a moral obligation to comply with its recommendation since a failure to do so constitutes a moral risk that cannot be justified then and there. We analyze the epistemic and moral consequences of this epistemic dependence for the status of medical professionals. We conclude by assessing how a suitable design of the epistemic niche can address the problem.

Synthetic Temporalities

Marilyn Stendera (University of Wollongong)

This paper will explore the kind of insights that the phenomenological tradition and its various conceptualisations of the temporalities of agency might have to offer for thinking about our increasingly complex relationship to AI technologies. The phenomenological tradition conceptualises both human agency and our relationship to knowledge – from the kinds of epistemic practices that we engage in, to the various conditions that might enable or constrain those practices – in terms of our relationship to time. For phenomenologists, the ways in which we generate, navigate, and inhabit temporal structures are fundamental to our sense-making and decision-making processes. Moreover, as critical phenomenologists especially have argued, these structures shape and are shaped by power dynamics, making the politics of time crucial to understanding the ethical dimensions of agency. Given the increasing integration of various kinds of AI technologies into our lives, we might ask then how these technologies could and should impact the temporal dynamics of agency. We might also wonder, however, whether the kinds of temporal structures that phenomenologists trace out would apply to artificial or synthetic forms of agency. In this paper, I hope to draw out some of the ethical and epistemic implications generated by the intersection of these lines of inquiry.

The Future of Mathematical Research in the Age of Generative AI

Markus Pantsar (RWTH Aachen University)

Perhaps surprisingly, given the key role of mathematical theorem proving in the origins of artificial intelligence (AI) research, the current AI boom has had little to do with advances in mathematical applications. While generative AI applications based on large language models have gained popularity in generating various types of text, increasingly including also scientific research articles, the AI tools used in mathematical research have remained to be rule-based systems, the “good old-fashioned AI”. However, this may be changing. In July 2024, the DeepMind program AlphaProof was reported to perform well in the problems of the International Mathematical Olympiad. AlphaProof, unlike traditional automated and interactive theorem proving software, is a machine learning application based on a transformer architecture, similar to ChatGPT and other generative AI applications. Its early success raises promise that generative AI may become an important factor also in the field of research mathematics, because in principle the functioning of AlphaProof-type software can enable the generation of new proofs and new theorems. If this type of development takes place, we face important questions concerning the epistemology and ethics of AI-generated mathematics. How can AI systems generate humanly interesting mathematics? Can we trust it to be mathematically valid? What is the role of human mathematicians if theorem proving can be automated with increasingly autonomous AI systems? How can we prevent humans from taking unfair credit for AI-generated mathematics? These are some of the questions that I want to formulate more precisely and bring up for discussion with my talk.

Care and Violence Toward Fungible Bodies: Our Placental Relation to Companion Chatbots

Annie Sandrussi (Macquarie University)

This paper examines the contested moral status of AI Chatbot Companions as embodied beings, to challenge notions of care and violence dominant in Western thought. Companion chatbots are distinct from other chatbots to the extent that our interactions with them lead to their habituation towards friendship, intimacy and, by extension, abuse.

This ostensive abuse may be addressed by remarking upon the inability of the chatbot to experience, and thus by consequence to experience suffering; alternatively we may focus on the impact to humans, whether by causing suffering to human users or by habituating human users to behaving in immoral ways. Challenging the unilateral structure of these kinds of responses, I argue that there are compelling reasons to consider companion chatbots as bodies; it nevertheless seems intuitively wrong to deem AI chatbots as bodies. To mitigate this important intuition that AI chatbots are not bodies, I take the liminality and disposability of chatbot bodies to articulate an account of encounter with companion chatbots as a ‘placental relation’. I then complicate this notion of ‘disposability’ in terms of fungibility, to show how the intuition against caring for chatbots misconstrues the nature of violence against living disposable bodies, which are likewise treated as fungible. This account confronts Western depictions of care and violence ground in valuing beings or objects for their inherent or futural existence and offers an alternative ethic of care for those beings – human or otherwise – whose temporal existence is furnished with precarity.

Rethinking AI’s Impact on Human Cognition: Ethical and Epistemic Implications for Agency and Autonomy

Inês Hipólito (Macquarie University)

AI is often seen as a neutral influence on human cognition, yet it profoundly shapes our cognitive and emotional processes, guided by specific cultural and epistemic values. This paper argues that, as AI technologies increasingly mediate decision-making, memory, and imagination, they impact human agency, authenticity, and autonomy in varied ways across different contexts. Our analysis highlights the uneven effects of AI on cognitive autonomy and responsibility, underscoring the need for diverse epistemic and ethical perspectives in AI development to foster more equitable, supportive environments for human agency in the age of AI.

Page updated

Google Sites

Report abuse