Atay Kozlovski

Digital duplication refers to the creation of interactive systems that simulate individuals, whether living or deceased and has become a widely discussed topic in the ethics of artificial intelligence. One central issue concerns the feasibility of consent to digital duplication, with some authors arguing that digital duplicates raise unique challenges for consent. Drawing on established debates in bioethics, this paper questions that uniqueness thesis by showing that the difficulties associated with digital duplication largely mirror well-known issues in the medical domain. Building on this analysis, it argues that the challenges posed by digital duplication do not genuinely undermine consent and presents consent models from the bioethical literature that are informative for the ethics of digital duplication.

AI generated simulations of real people, also labelled Digital Duplicates (DDs), are increasingly deployed across personal, educational, professional and heritage contexts. Such systems necessarily involve some notion of accuracy in simulating the target person and aiming for a maximally accurate simulation seems to be a natural design approach that is commonly presupposed across the industry and the philosophical literature on the topic. We challenge this approach and argue for deliberate choices about the level and kind of accuracy a DD should aim for. This article analyses what it means to accurately simulate a person, introduces a typology of five categories of inaccuracy likely to arise in DDs, and discusses their ethical implications. Applying this typology to the case of griefbots, the paper goes on to illustrate how the different categories interact in practice and that the desirable degree and type of accuracy is highly context-dependent.

Advances in generative artificial intelligence (AI) have driven a growing effort to create digital duplicates. These semi-autonomous recreations of living and dead people can be used for many purposes. Some of these purposes include tutoring, coping with grief, and attending business meetings. However, the normative implications of digital duplicates remain obscure, particularly considering the possibility of them being applied to genocide memory and education. To address this gap, we examine normative possibilities and risks associated with the use of more advanced forms of generative AI-enhanced duplicates for transmitting Holocaust survivor testimonies. We first review the historical and contemporary uses of survivor testimonies. Then, we scrutinize the possible benefits of using digital duplicates in this context and apply the Minimally Viable Permissibility Principle (MVPP). The MVPP is an analytical framework for evaluating the risks of digital duplicates. It includes five core components: the need for authentic presence, consent, positive value, transparency, and harm-risk mitigation. Using MVPP, we identify potential harms digital duplicates might pose to different actors, including survivors, users, and developers. We also propose technical and socio-technical mitigation strategies to address these harms.

Recent developments in theories for responsible innovation have focused on the importance of actively accounting for values in our technological designs. Leading among these theories is that of Value Sensitive Design which attempts to guide the design process on the basis of evaluative analysis. However, values often come into conflict and VSD has been criticized for not providing a proper method to resolve such inevitable conflicts. This paper examines three such methods and argues that although each has its merits, they all fail to account for a common source of value conflicts known as value incommensurability. Drawing on literature from the field of axiology, this paper argues that by incorporating the evaluative relation of ‘parity’ each of these three methods, and the VSD framework in general, will be able to properly understand the relation which holds between conflicting design options stemming from the incommensurable of values and be able to guide designers in making rational decision in the face of such conflicts.

The rapid proliferation of AI systems has raised many concerns about safety and responsibility in their design and use. The philosophical framework of Meaningful Human Control (MHC) was developed in response to these concerns, and tries to provide a standard for designing and evaluating such systems. While promising, the framework still requires further theoretical and practical refinement. This paper contributes to that effort by drawing on research in axiology and rational decision theory to identify a critical gap in the framework. Specifically, it argues that while ‘reasons’ play a central role in MHC, there has been little discussion of the possibility that, when weighed against each other, reasons may not always point to a single, rationally preferable course of action. I refer to these cases as instances of reasons underdetermination, and this paper discusses the need to address this issue within the MHC framework. The paper begins by providing an overview of the key concepts of the MHC framework and then examines the role of ‘reasons’ in the framework’s two main conditions - Tracking and Tracing. It then discusses the phenomenon of reasons underdetermination and shows how it poses a challenge for the achievement of both Tracking and Tracing.

Advances in generative AI have given rise to a growing industry centred on interactive representations of deceased individuals. Within this emerging “digital afterlife industry”, interactive deadbots (IDBs) are presented as hyper-realistic avatars that use a person’s likeness, voice, and personal data to simulate conversational interactions with them. Rapidly moving from a niche experiment to a mainstream phenomenon, IDBs are poised to reshape the ethical, social, legal, and governance landscapes surrounding death, mourning, and digital legacy. This paper examines the disruptive nature of IDB technology through a multidisciplinary lens, using the concept of indeterminacy as its guiding analytical framework and a novel way to conceptualise the unstable field. Rather than advancing a unified understanding of indeterminacy, we introduce a structured analytical map and provisional taxonomy that distinguishes technological, social, philosophical, legal, and regulatory manifestations of indeterminacy in IDBs. By offering a tentative and necessarily selective map of this fluid and nascent field, we explore how indeterminacy and IDBs intersect. The paper examines how IDBs amplify existing forms of indeterminacy and how indeterminacy itself shapes the development and use of these systems across five domains: technological, social, philosophical, legal, and regulatory.

Digital duplicates (DDs), or artificial intelligence (AI) simulations of individual minds, are AI models designed to emulate real people. Depending on how they are trained, DDs can reproduce aspects or signs of a person’s psychology, values and communication style, enabling interaction with digital approximations of living, recently deceased or historical individuals. Proposed uses include education, advocacy, grief support, medical decision-­ making and the continuation of creative or personal projects. While potentially valuable, these applications also pose ethical risks such as privacy loss, misrepresentation, manipulation and the erosion of authentic human relationships. This paper argues that DDs should be treated as context-­ specific tools rather than general-­ purpose technologies. We outline a two-­ tier ethical framework grounded in considerations of harm and proportionality, supported by the values of information permission, transparency, authenticity and appropriate access. The framework unites diverse moral concerns under a single framework while allowing their relative weight to vary across domains such as medicine, education, memorialisation and entertainment. By situating ethical evaluation within a harm-­ reduction approach that is sensitive to context, we offer a practical basis for the responsible development and governance of human-emulating AI systems.