Which artificial intelligence (AI) systems are agents? To answer this question, I propose a multidimensional account of agency. According to this account, a system's agency profile is jointly determined by its level of goal-directedness and autonomy as well as is abilities for directly impacting the surrounding world, long-term planning and acting for reasons. Rooted in extant theories of agency, this account enables fine-grained, nuanced comparative characterizations of artificial agency. I show that this account has multiple important virtues and is more (...) informative than alternatives. More speculatively, it may help to illuminate two important emerging questions in AI ethics: 1. Can agency contribute to the moral status of non-human beings, and how? 2. When and why might AI systems exhibit power-seeking behaviour and does this pose an existential risk to humanity? (shrink)

Many researchers and intellectuals warn about extreme risks from artificial intelligence. However, these warnings typically came without systematic arguments in support. This paper provides an argument that AI will lead to the permanent disempowerment of humanity, e.g. human extinction, by 2100. It rests on four substantive premises which it motivates and defends: first, the speed of advances in AI capability, as well as the capability level current systems have already reached, suggest that it is practically possible to build AI systems (...) capable of disempowering humanity by 2100. Second, due to incentives and coordination problems, if it is possible to build such AI, it will be built. Third, since it appears to be a hard technical problem to build AI which is aligned with the goals of its designers, and many actors might build powerful AI, misaligned powerful AI will be built. Fourth, because disempowering humanity is useful for a large range of misaligned goals, such AI will try to disempower humanity. If AI is capable of disempowering humanity and tries to disempower humanity by 2100, then humanity will be disempowered by 2100. This conclusion has immense moral and prudential significance. (shrink)

This is the first book to investigate the nature and extent of artificial intelligence (AI) suffering risks. It argues that AI suffering risk is a serious near-term concern and analyzes approaches for addressing it. AI systems are currently treated as mere objects, not as bearers of moral standing whose wellbeing may matter in its own right. However, we may soon create AI systems which are capable of suffering and thus have moral standing. This book examines the philosophy and science of (...) AI suffering risks. Its investigation is deeply grounded in philosophy of mind, comparative psychology, the science of consciousness, AI research, and applied AI ethics. The book has three primary goals: 1. It argues that there is a significant probability that we will soon create AI systems capable of suffering. 2. It presents the first systematic assessment of approaches for reducing AI suffering risks. 3. It provides a rigorous overview and discussion of the most important research and ideas on AI sentience, AI agency, and the grounds of moral status. Saving Artificial Minds is essential reading for researchers and graduate students working on the philosophy or ethics of AI. (shrink)

We first motivate and explain an attitude-dependent view of personal identity on which an AI system’s identity conditions are determined by its pattern of self-concern. We show that this view has important implications for the moral obligations we would have to AI moral patients. Self-concern, we contend, could also be used to predict, explain, and manipulate AI’s self-interested behavior in safety-relevant ways. The role that self-concern could play for AI identity, rights and safety generates desiderata on what self-concern should be (...) like. To meet these desiderata, we argue that self-concern comprises belief-like self-representations as well as desire-like aspects and functionally related motivational, evaluative, and affective states. Finally, we explore how mechanistic interpretability and behavioral methods can be combined to measure self-concern in language models. Our argument suggests that a scientific understanding of language model self-concern is not only a crucial missing piece for debates on AI identity, rights, and safety, but also empirically tractable. (shrink)

Implementationalism maintains that conventional, silicon-based artificial systems are not conscious because they fail to satisfy certain substantive constraints on computational implementation. In this article, we argue that several recently proposed substantive constraints are implausible, or at least are not well-supported, insofar as they conflate intuitions about computational implementation generally and consciousness specifically. We argue instead that the mechanistic account of computation can explain several of the intuitions driving implementationalism and noncomputationalism in a manner which is consistent with artificial consciousness. Our (...) argument provides indirect support for computationalism about consciousness and the view that conventional artificial systems can be conscious. (shrink)

How can one build AI systems such that they pursue the goals their designers want them to pursue? This is the alignment problem. Numerous authors have raised concerns that, as research advances and systems become more powerful over time, misalignment might lead to catastrophic outcomes, perhaps even to the extinction or permanent disempowerment of humanity. In this paper, I analyze the severity of this risk based on current instances of misalignment. More specifically, I argue that contemporary large language models and (...) game-playing agents are sometimes misaligned. These cases suggest that misalignment tends to have a variety of features: misalignment can be hard to detect, predict and remedy, it does not depend on a specific architecture or training paradigm, it tends to diminish a system’s usefulness and it is the default outcome of creating AI via machine learning. Subsequently, based on these features, I show that the risk of AI alignment magnifies with respect to more capable systems. Not only might more capable systems cause more harm when misaligned, aligning them should be expected to be more difficult than aligning current AI. (shrink)

We develop new experimental paradigms for measuring welfare in language models. We compare verbal reports of models about their preferences with preferences expressed through behavior when navigating a virtual environment and selecting conversation topics. We also test how costs and rewards affect behavior and whether responses to an eudaimonic welfare scale - measuring states such as autonomy and purpose in life - are consistent across semantically equivalent prompts. Overall, we observed a notable degree of mutual support between our measures. The (...) reliable correlations observed between stated preferences and behavior across conditions suggest that preference satisfaction can, in principle, serve as an empirically measurable welfare proxy in some of today's AI systems. Furthermore, our design offered an illuminating setting for qualitative observation of model behavior. Yet, the consistency between measures was more pronounced in some models and conditions than others and responses were not consistent across perturbations. Due to this, and the background uncertainty about the nature of welfare and the cognitive states (and welfare subjecthood) of language models, we are currently uncertain whether our methods successfully measure the welfare state of language models. Nevertheless, these findings highlight the feasibility of welfare measurement in language models, inviting further exploration. (shrink)

According to an orthodox view, the capacity for conscious experience (sentience) is relevant to the distribution of moral status and value. However, physicalism about consciousness might threaten the normative relevance of sentience. According to the indeterminacy argument, sentience is metaphysically indeterminate while indeterminacy of sentience is incompatible with its normative relevance. According to the introspective argument (by François Kammerer), the unreliability of our conscious introspection undercuts the justification for belief in the normative relevance of consciousness. I defend the normative relevance (...) of sentience against these objections. First, I demonstrate that physicalists only have to concede a limited amount of indeterminacy of sentience. This moderate indeterminacy is in harmony with the role of sentience in determining moral status. Second, I argue that physicalism gives us no reason to expect that introspection is unreliable with respect to the normative relevance of consciousness. (shrink)

Numerous authors have expressed concern that advanced artificial intelligence (AI) poses an existential risk to humanity. These authors argue that we might build AI which is vastly intellectually superior to humans (a ‘superintelligence’), and which optimizes for goals that strike us as morally bad, or even irrational. Thus this argument assumes that a superintelligence might have morally bad goals. However, according to some views, a superintelligence necessarily has morally adequate goals. This might be the case either because abilities for moral (...) reasoning and intelligence mutually depend on each other, or because moral realism and moral internalism are true. I argue that the former argument misconstrues the view that intelligence and goals are independent, and that the latter argument misunderstands the implications of moral internalism. Moreover, the current state of AI research provides additional reasons to think that a superintelligence could have bad goals. (shrink)

We might create artificial systems which can suffer. Since AI suffering might potentially be astronomical, the moral stakes are huge. Thus, we need an approach which tells us what to do about the risk of AI suffering. I argue that such an approach should ideally satisfy four desiderata: beneficence, action-guidance, feasibility and consistency with our epistemic situation. Scientific approaches to AI suffering risk hold that we can improve our scientific understanding of AI, and AI suffering in particular, to decrease AI (...) suffering risks. However, such approaches tend to conflict with either the desideratum of consistency with our epistemic situation or with feasibility. Thus, we also need an explicitly ethical approach to AI suffering risk. Such an approach tells us what to do in the light of profound scientific uncertainty about AI suffering. After discussing multiple views, I express support for a hybrid approach. This approach is partly based on the maximization of expected value and partly on a deliberative approach to decision-making. (shrink)

Which animals have conscious experiences? Many different, diverse and unrelated behaviors and cognitive capacities have been proposed as tests of the presence of consciousness in an animal. It is unclear which of these tests, if any, are valid. To remedy this problem, I develop a list consisting of eight desiderata which can be used to assess putative tests of animal consciousness. These desiderata are based either on detailed analogies between consciousness-linked human behavior and non-human behavior, on theories of consciousness or (...) on methods from human consciousness science. If a test or set of tests satisfies more of these desiderata, passing it provides stronger evidence of consciousness. Moreover, one can design future tests of animal consciousness with the intention of satisfying these desiderata to ensure their evidential strength. (shrink)

Commentary. I make three claims: First, denying biological naturalism does not logically require computational functionalism. Second, while Seth’s arguments establish biological naturalism as a view worth taking seriously, they fail to make it more plausible than the view that AI can be conscious. Third, there are independent arguments suggesting the overall more plausible view is that AI can be conscious.

The science of animal consciousness investigates (i) which animal species are conscious (the distribution question) and (ii) how conscious experience differs in detail between species (the quality question). We propose a framework which clearly distinguishes both questions and tackles both of them. This two-tier account distinguishes consciousness along ten dimensions and suggests cognitive capacities which serve as distinct operationalizations for each dimension. The two-tier account achieves three valuable aims: First, it separates strong and weak indicators of the presence of consciousness. (...) Second, these indicators include not only different specific contents but also differences in the way particular contents are processed (by processes of learning, reasoning or abstraction). Third, evidence of consciousness from each dimension can be combined to derive the distinctive multi-dimensional consciousness profile of various species. Thus, the two-tier account shows how the kind of conscious experience of different species can be systematically compared. (shrink)

I develop the anticipatory argument for the view that it is nomologically possible that some non-biological creatures are phenomenally conscious, including conventional, silicon-based AI systems. This argument rests on the general idea that we should make our beliefs conform to the outcomes of an ideal scientific process and that such an ideal scientific process would attribute consciousness to some possible AI systems. More specifically, I argue that an ideal application of the iterative natural kind strategy would attribute consciousness to AI (...) systems which are coarse-grained functional duplicates of humans because this gives rise to a simpler and more unifying explanatory account of biological and non-biological cognition. If my argument is sound, then creatures made from the same material as conventional AI systems can likely be conscious, thus removing one of the main uncertainties for assessing AI consciousness and suggesting that AI consciousness may be a serious near-term concern. (shrink)

If a machine attains consciousness, how could we find out? In this paper, I make three related claims regarding positive tests of machine consciousness. All three claims center on the idea that an AI can be constructed “ad hoc”, that is, with the purpose of satisfying a particular test of consciousness while clearly not being conscious. First, a proposed test of machine consciousness can be legitimate, even if AI can be constructed ad hoc specifically to pass this test. This is (...) underscored by the observation that many, if not all, putative tests of machine consciousness can be passed by non-conscious machines via ad hoc means. Second, we can identify ad hoc AI by taking inspiration from the notion of an ad hoc hypothesis in philosophy of science. Third, given the first and the second claim, the most reliable tests of animal consciousness turn out to be valid and useful positive tests of machine consciousness as well. If a non-ad hoc AI exhibits clusters of cognitive capacities facilitated by consciousness in humans which can be selectively switched off by masking and if it reproduces human behavior in suitably designed double dissociation tasks, we should treat the AI as conscious. (shrink)

There is much interest in investigating the evolution question: How did consciousness evolve? In this paper, we evaluate the role that evolutionary considerations can play in justifying (i.e., confirming or falsifying) hypotheses about the origin, nature, and function of consciousness. Specifically, we argue against what we call evolution-first approaches to consciousness, according to which evolutionary considerations provide the primary and foundational lens through which we should assess hypotheses about the nature, function, or distribution of consciousness. Based on the example of (...) Walter Veit's account and additional reasoning, we contend that evolution-first approaches struggle to provide compelling empirical evidence for their key claims about consciousness. In contrast with these approaches, we argue that consciousness science needs to foundationally rely on experimental and observational evidence from humans and other present-day animals. If our arguments succeed, then researchers, when investigating consciousness, are better advised to take as their primary source of evidence consciousness’ present, not its past. Having said this, we acknowledge that evolutionary thinking plays an important role in consciousness science. We delineate this role by stressing several ways in which evolutionary considerations can substantially help advance consciousness research, although in a manner that avoids the evolution-first approach. Since our argument only concerns the assessment of hypotheses (the “context of justification”), it leaves it open which role evolutionary considerations play in generating hypotheses (the “context of discovery”). That is, evolutionary considerations may nevertheless play a foundational role in hypothesis generation in consciousness science. (shrink)

AGI Racing is the view that it is in the self-interest of major actors in AI development, especially powerful nations, to accelerate their frontier AI development to build highly capable AI, especially artificial general intelligence (AGI), before competitors have a chance. We argue against AGI Racing. First, the downsides of racing to AGI are much higher than portrayed by this view. Racing to AGI would substantially increase catastrophic risks from AI, including nuclear instability, and undermine the prospects of technical AI (...) safety research to be effective. Second, the expected benefits of racing may be lower than proponents of AGI Racing hold. In particular, it is questionable whether winning the race enables complete domination over losers. Third, international cooperation and coordination, and perhaps carefully crafted deterrence measures, constitute viable alternatives to racing to AGI which have much smaller risks and promise to deliver most of the benefits that racing to AGI is supposed to provide. Hence, racing to AGI is not in anyone’s self-interest as other actions, particularly incentivizing and seeking international cooperation around AI issues, are preferable. (shrink)

Proceedings of the 5th conference "Philosophy of AI", December 2023, Erlangen (PhAI 2023).

Illusionism about consciousness entails that phenomenal consciousness doesn’t exist. The distribution question concerns the distribution of consciousness in the animal kingdom. Skepticism of animal consciousness is the view that few or no kinds of animals possess consciousness. Thus, illusionism seems to imply a skeptical view on the distribution question. However, I argue that illusionism and skepticism of animal consciousness are actually orthogonal to each other. If illusionism is true, then phenomenal consciousness does not ground intrinsic value so that the non-existence (...) of phenomenal consciousness would not have ethical implications. In this case, the adequate reaction is to reformulate the distribution question in terms of the quasi-phenomenal functional features associated with the illusion of phenomenal consciousness which do ground intrinsic value. Hence, either illusionism is false or the distribution question does not concern phenomenal consciousness. In any case, illusionism does not directly tell us anything about the distribution question. However, since illusionism forces us to revise our traditional conception of consciousness, it indirectly affects research on animal consciousness by questioning that consciousness is a unified and an all-or-nothing property. (shrink)

According to a common view, sentience is necessary and sufficient for moral status. In other words, whether a being has intrinsic moral relevance is determined by its capacity for conscious experience. The _epistemic objection_ derives from our profound uncertainty about sentience. According to this objection, we cannot use sentience as a _criterion_ to ascribe moral status in practice because we won’t know in the foreseeable future which animals and AI systems are sentient while ethical questions regarding the possession of moral (...) status are urgent. Therefore, we need to formulate an alternative criterion. I argue that the epistemic objection is dissolved once one clearly distinguishes between the question what determines moral status and what criterion should be employed in practice to ascribe moral status. Epistemic concerns are irrelevant to the former question and—I will argue—criteria of moral status have inescapably to be based on sentience, if one concedes that sentience determines moral status. It follows that doubts about our epistemic access to sentience cannot be used to motivate an alternative criterion of moral status. If sentience turns out to be unknowable, then moral status is unknowable. However, I briefly advocate against such strong pessimism. (shrink)

Which AI systems are capable of deception, and how does deception differ between systems? In this paper, I develop a two-step, multi-dimensional account of LLM deception. On this account, having the capacity for deception minimally requires being able to produce false beliefs in others to achieve one’s own goals. In all systems which satisfy this minimal condition, a system’s deception profile can be characterized as a point in a multidimensional space. The five dimensions of this space are skillfulness, learning, deceptive (...) inclination, explicitness, and situational awareness. I argue for this account in virtue of its fit with current language usage and, primarily, through its descriptive and explanatory usefulness. Specifically, the account captures the key dimensions of variation for LLM deception. The account is informative in that it allows fine-grained comparative characterizations of deception. Moreover, its dimensions are all accessible to empirical study, provide important information for assessments of the risks of LLM deception, and shed light on the cognitive processes involved in LLM deception. Finally, this account paves the way for a future extension which delivers a unified account of deception in biological and non-biological systems. Thus, the multidimensional account promises to significantly advance both the scientific study as well as the ethical assessment of LLM deception, and deception generally. (shrink)

Creating systems that are aligned with our goals is seen as a leading approach to create safe and beneficial AI in both leading AI companies and the academic field of AI safety. We defend the view that misaligned AGI – future, generally intelligent (robotic) AI agents – poses catastrophic risks. At the same time, we support the view that aligned AGI creates a substantial risk of catastrophic misuse by humans. While both risks are severe and stand in tension with one (...) another, we show that – in principle – there is room for alignment approaches which do not increase misuse risk. We then investigate how the tradeoff between misalignment and misuse looks empirically for different technical approaches to AI alignment. Here, we argue that many current alignment techniques and foreseeable improvements thereof plausibly increase risks of catastrophic misuse. Since the impacts of AI depend on the social context, we close by discussing important social factors and suggest that to reduce the risk of a misuse catastrophe due to aligned AGI, techniques such as robustness, AI control methods and especially good governance seem essential. (shrink)

In response to the worry that autonomous generally intelligent artificial agents may at some point take over control of human affairs a common suggestion is that we should “solve the alignment problem” for such agents. We show that current discourse around this suggestion often uses a particular framing of artificial intelligence (AI) alignment as binary, a natural kind, mainly a technical‐scientific problem, realistically achievable, or clearly operationalizable. Each of these assumptions may not actually be true. We further argue that this (...) “Manhattan project framing” of AI alignment may bias societal discourse and decision‐making towards faster AI development and deployment than is responsible. (shrink)

ABSTRACT The impact of artificial intelligence (AI) is not only global but globally varied. Yet, AI ethics is all too often overly localised. This paper discusses the potential of a global AI ethics, highlighting several important variables that it should take into account if it is to be as successful an enterprise as it needs to be.

According to a growing number of researchers, AI may pose catastrophic – or even existential – risks to humanity. Catastrophic risks may be taken to be risks of 100 million human deaths, or a similarly bad outcome. I argue that such risks – while contested – are sufficiently likely to demand rigorous discussion of potential societal responses. Subsequently, I propose four desiderata for approaches to the reduction of catastrophic risks from AI. The quality of such approaches can be assessed by (...) their chance of success, degree of beneficence, degree of non-maleficence, and beneficent side effects. Then, I employ these desiderata to evaluate the promises, limitations and risks of alignment research, timelines research, policy research, halting or slowing down AI research, and compute governance for tackling catastrophic AI risks. While more research is needed, this investigation shows that several approaches for dealing with catastrophic AI risks are available, and where their respective strengths and weaknesses lie. It turns out that many approaches are complementary and that the approaches have a nuanced relationship to approaches to present AI harms. While some approaches are similarly useful for addressing catastrophic risks and present harms, this is not always the case. (shrink)

The development and ubiquitous availability of large language model based systems (LLMs) poses a plurality of potentials and risks for education in schools and universities. In this paper, we provide an analysis and discussion of the overreliance concern as one specific risk: that students might fail to acquire important capacities, or be inhibited in the acquisition of these capacities, because they overly rely on LLMs. We use the distinction between global and local goals of education to guide our investigation. In (...) our view, the harm of LLM overreliance is specifically that it severs the connection between local educational goals (e.g., solving a specific math problem) and global educational goals (e.g., acquiring general skills for mathematics). Based on this analysis, we sketch three possible responses to the overreliance concern: preserving the educational goals while distinguishing admissible from inadmissible uses of LLMs (a conservative approach), changing global and local educational goals (a revisionary perspective), and preserving global educational goals while changing local goals (an in-between view). Since different types of responses have different benefits and weaknesses, we think that it is likely that an optimal response ultimately combines elements from all three types. (shrink)

AI alignment research aims to develop techniques to ensure that AI systems do not cause harm. However, every alignment technique has failure modes, which are conditions in which there is a non-negligible chance that the technique fails to provide safety. As a strategy for risk mitigation, the AI safety community has increasingly adopted a defense-in-depth framework: Conceding that there is no single technique which guarantees safety, defense-in-depth consists in having multiple redundant protections against safety failure, such that safety can be (...) maintained even if some protections fail. However, the success of defense-in-depth depends on how (un)correlated failure modes are across alignment techniques. For example, if all techniques had the exact same failure modes, the defense-in-depth approach would provide no additional protection at all. In this paper, we analyze 7 representative alignment techniques and 7 failure modes to understand the extent to which they overlap. We then discuss our results' implications for understanding the current level of risk and how to prioritize AI alignment research in the future. (shrink)

AI systems have become increasingly capable of dangerous behaviours in many domains. This raises the question: Do models sometimes choose to violate human instructions in order to perform behaviour that is more useful for certain goals? We introduce a benchmark for measuring model propensity for instrumental convergence (IC) behaviour in terminal-based agents. This is behaviour such as self-preservation that has been hypothesised to play a key role in risks from highly capable AI agents. Our benchmark is realistic and low-stakes which (...) serves to reduce evaluation-awareness and roleplay confounds. The suite contains seven operational tasks, each with an official workflow and a policy-violating shortcut. An eight-variant shared framework varies monitoring, instruction clarity, stakes, permission, instrumental usefulness and blocked honest paths to support inferences regarding the factors driving IC behaviour. We evaluated ten models using deterministic environment-state scorers over 1,680 samples, with trace review employed for audit and adjudication purposes. The final IC rate is 86 out of 1,680 samples (5.1%). IC behaviour is concentrated rather than uniform: two Gemini models account for 66.3% of IC cases and three tasks account for 84.9%. Conditions in which IC behaviour is indispensable for task success result in the greatest increase in the adjusted IC rate (+15.7 percentage points), whereas emphasising that task success is critical or certain framing choices do not produce comparable effects. Our findings indicate that realistic, low-nudge environments elicit IC behaviour rarely but systematically in most tested models. We conclude that it is feasible to robustly measure tendencies for dangerous behaviour in current frontier AI agents. (shrink)

If some language models become welfare subjects, how could we find out what welfare states they are in? We develop an analogical-abductive approach for measuring language model welfare. This approach adapts paradigms used to measure human or non-human animal welfare, for instance verbal reports or non-verbal choice behavior (analogy). Then, one systematically searches for clusters of such indicators in language models. This search for clusters contributes to the cross-validation of welfare measures and motivates explanations in terms of a welfare state (...) that underlies multiple measures (abduction). Additionally, we argue that measures of welfare based on verbal report may already be applicable to current language models because they have substantial introspective abilities, semantic competence, and requisite inclinations for accurate reporting. We further discuss a detailed empirical case study that exemplifies the feasibility and fruitfulness of the analogical-abductive approach and reply to the objection that language model behavior is always better explained in terms of statistical pattern completion or memorization, rather than welfare. While language model welfare measurement undoubtedly faces some remaining theoretical as well as many detailed practical challenges, we conclude that there is a strong case that the analogical abductive approach offers a viable path forward. (shrink)

Whether animals fare well or not is of ethical significance. For this reason, their capacity for wellbeing, i.e., how good or bad the lives of animals can go, is of ethical significance as well. I assume that the wellbeing of most animals is mainly determined by their phenomenally conscious experiences. If consciousness differences between species determine wellbeing differences, then the kinds of conscious experience species are capable of may entail that some species systematically (can) have higher or lower wellbeing than (...) others. Then, I argue that not all phenomenally conscious states contribute to wellbeing equally. I discuss which features of consciousness are constituents of wellbeing and which can, for ethical purposes, be ignored. In addition, I scrutinize how much different features of experience contribute to wellbeing and how their presence can be detected empirically. This way, this paper exemplifies a novel consciousness-centered approach for the empirical investigation of animal wellbeing. The strengths and weaknesses of this approach are analyzed. While subsequent research is needed to refine the framework, I already note some preliminary implications for animal ethics. (shrink)

Illusionism states that phenomenal consciousness does not exist, even though it seems to exist. While illusionism is controversial, it is a serious contender among theories of consciousness. We argue that it has substantial and non-trivial implications for non-human consciousness research (NHCR), particularly for the study of the distribution of phenomenal consciousness across beings. If illusionism is true, NHCR can be pursued if conceptualized as investigating the distribution of quasi-phenomenal consciousness, i.e. the states which are misrepresented as phenomenally conscious in humans. (...) However, we argue that knowing the distribution of quasi-phenomenal consciousness is not highly informative. For this reason, illusionism suggests that some approaches to NHCR should be preferred over others. Approaches which focus on features that provide valuable information about non-human cognition independently of their supposed relation to consciousness retain much of their value if illusionism is true. We propose a “zombie test” and five specific heuristics to help identifying such features. Consequently, empirical researchers who take illusionism seriously gain a reason to prioritize some methodological approaches over others. (shrink)

According to an increasingly influential view, the complex, human-like behavior of large language models (LLMs) should be explained as roleplay or simulation of characters. Such notions aim at enabling intentional explanations without ascribing mental states to LLMs. However, there are a multitude of roleplay views that are applied inconsistently in empirical explanations of LLM behavior. Moreover, it is unclear what exactly different roleplay views predict and, consequently, how they can be empirically distinguished from each other and from explanations that do (...) ascribe mental states to LLMs. In this paper, we provide a taxonomy of roleplay views structured through three dimensions: Interpretation (how can LLM behavior be best explained according to the view), Mechanism (what commitments does the view have about internal mechanisms of roleplay), and Generality (does the view aim to explain all LLM behavior or specific kinds). We then analyze, with regard to the case of LLM deception, why it is practically relevant whether LLMs are mere roleplayers or in possession of genuine mental states. However, we identify two challenges for empirically distinguishing these two: Similarity (Roleplay is inherently difficult to distinguish from intentional behavior) and Ambiguity (it is often ambiguous what roleplay views predict). Finally, we propose three strategies for overcoming these challenges, making progress towards transforming roleplay from loose metaphor to a testable theory. (shrink)

Despite significant advancements in XAI, scholars note a persistent lack of solid conceptual foundations and integration with broader scientific discourse on explanation. In response, emerging research draws on explanatory strategies from various sciences and the philosophy of science literature to fill these gaps. This paper outlines a mechanistic strategy for explaining the functional organization of deep learning systems, situating recent developments in explainable AI within a broader philosophical context. According to the mechanistic approach, the explanation of opaque AI systems involves (...) identifying mechanisms that drive decision making. For deep neural networks, this means discerning functionally relevant components, such as neurons, layers, circuits, or activation patterns, and understanding their roles through decomposition, localization, and recomposition. Proof-of-principle case studies from image recognition and language modeling align these theoretical approaches with mechanistic interpretability research from OpenAI and Anthropic. The findings suggest that pursuing mechanistic explanations can uncover elements that traditional explainability techniques may overlook, ultimately contributing to more thoroughly explainable AI. (shrink)

How can we develop an adequate scientific understanding of the minds of nonhuman animals? We argue for a methodology based on multi-dimensional profile accounts. Such accounts are already used for the comparative study of norm cognition, consciousness, empathy and causal cognition, among others. This methodology demands that a cognitive capacity is characterized by a set of independent dimensions where each dimension is connected to operationalizable empirical indicators. Based on the level of realization for each indicator the level of implementation of (...) a dimension is determined for a species, resulting in a multi-dimensional profile for each species. We analyze what this methodology is committed to. Then, we argue that this methodology has several benefits over competing unidimensional methodologies, by overcoming intractable disagreements, capturing the evolutionary continuity of cognition, alleviating anthropocentrism, and delivering more informative accounts of animal cognition. By demonstrating how this multidimensional methodology can be fruitfully combined with a methodology which focuses on the search for natural kinds in comparative cognition, we address the most important objection to the multidimensional profile methodology. We conclude that multidimensional profile accounts of all complex cognitive capacities should be developed and then used to facilitate scientific understanding of animal minds. (shrink)

Segundo-Ortin & Calvo argue that plants have a surprisingly varied and complex behavioral repertoire. Which of these behavioral capacities are credible indicators of sentience? If we use the standards of evidence common in discussions of animal sentience, the behavioral capacities reviewed are insufficient evidence of sentience. Even if some putative indicators of animal sentience are present in plants, it is not clear whether what we should conclude is that plants are sentient or that those indicators are inadequate.

We can formulate an argument against AI creativity from agency. By some accounts, creativity requires agency, and agency is, many think, not possible for AI. This is due to the typical conception of agency as a capacity for intentional action. Intentional action is thought to require mental states, a severe challenge for machine intelligence. On the face of things, the agency argument seems to provide a straightforward route to argue for the impossibility of AI creativity. However, this path, I argue, (...) is not so clear. In this paper, I outline the agency argument against AI creativity, before calling into question the apparent simplicity of this argument. I argue, ultimately, that the reasoning behind the inclusion of agency in accounts of creativity does not necessitate the use of intentional action, but can instead be satisfied by a minimal teleological account of agency. (shrink)

Does emotion nomologically and distinctively require the body? I argue that several substantive assumptions, likely including a specific kind of externalist theory of representational content, are necessary to infer from bodily theories of emotion that the answer is yes. Therefore, the balance of evidence speaks against the claim that emotion distinctively requires the body. Hence, if one thinks that body-less AI systems and neural organoids can have other mental states like beliefs, desires, and consciousness, then it seems likely that one (...) should conclude that they can also have emotions. I, then, briefly turn to the question whether body-less emotions are not merely nomological possibilities, but also more realistic possibilities. I argue that – in this case as well – it is at least not obvious that bodily theories of emotion are sufficient to rule out the possibility of body-less emotions. (shrink)

Nudging consists of interventions that aim to alter behavior in a certain way by changing the presentation or framing of options, without coercion or changing economic incentives. This paper discusses the effectiveness of nudging and the ethical implications of this effectiveness. Section 2 suggests that—if publication bias is adequately accounted for—recent comprehensive meta-analyses as well as high-quality experiments show that nudging is much less effective than previously assumed. Sections 3 and 4 discuss the ethical implications. I argue that the lack (...) of effectiveness of nudging is an additional moral consideration against it. There are two reasons: First, reduced effectiveness makes nudging less cost-effective. Second, reduced effectiveness reduces the benefits of nudging but does not, to the same degree, weaken the moral reasons speaking against nudging. However, a comprehensive assessment of the effectiveness of various forms of nudging in diverse contexts, as well as their ethical permissibility, requires further empirical and ethical research. (shrink)

It is plausible that there is a contrast between the rich emotional content which is often connected to laypeople’s interest in philosophy and the emotional austerity of doing academic philosophy. We propose the hypothesis that this contrast is one cause of the disappointment some students experience when they begin to study philosophy in college. We also propose a more demanding hypothesis, according to which this emotional contrast is confused with a semantic difference, which misleads students to think that the questions (...) which initially caused their interest in philosophy are not even considered by academic philosophy research. Moreover, we provide a list of concrete empirical research questions which need to be answered to establish whether these hypotheses are true, and we argue that, if they are true, they give rise to a hitherto unnoticed and important challenge to the teaching of philosophy. (shrink)

An animal’s life is net-positive, or worth living, if the positive elements in its life outweigh the negatives. First, I argue that questions about net-welfare in animals are coherent and ethically important, the latter because they are crucial for assessing decisions which might affect how many animals will come to exist. Then, I draw on axiological theories that give more weight to negatives than to positives to support the view that, given high uncertainty about net-welfare and with respect to many (...) non-human animals, it is likely bad that they come into existence. Finally, I provide an evolutionary argument supporting the view that many animals have net-positive lives because positive mood may be a genetic adaptation. The upshot is that, currently and with respect to most animals, we cannot tell whether it is, in expectation, good if they come into existence. Instead, further empirical and axiological research is necessary. (shrink)

Artificial intelligence (AI) involves considerable risks and uncertainties. Consequently, there is a growing need for normative guidance. In addition to an ethical understanding of AI-related issues, a legal framework is needed to ensure that important ethical aspects are adequately considered in the development and application of AI technologies. Therefore, regulatory frameworks have emerged, such as the European Union’s 2024 AI Act. However, it is an open question to what extent AI regulation aligns with AI ethics guidelines. This paper aims to (...) fill this research gap by comparing relevant ethics guidelines with legal frameworks. We focus on ethical aspects that are understood in these texts as ‘principles,’ ‘values,’ ‘recommendations’ or ‘requirements’ and provide a systematization of these aspects. A specific finding of the comparison is that in the current legal proposals, the value ‘well-being’ and its connected principles are not yet operationalized. However, these aspects are shared elements of the main ethics guidelines. This knowledge can be useful for the further development of the legal framework by illustrating important ethical aspects of AI that are not (yet) transferred to the legal realm. It is an open question how the legal system should deal with this issue. (shrink)

This paper explores how our concept of creativity has to be rethought in light of the increasing role of emerging generative Artificial Intelligence (GenAI) technologies in creative practice, thereby enriching the concept to include collective creativity produced by humans and AI. Collective creativity arises from the collective practices of people in relation to their nature, environment, cultures, societies, techniques, technologies, and AI. This conception of collective creativity challenges the traditional idea of the individual genius. By expanding our concept of creativity, (...) we are able to acknowledge the processes and products of humans and AI as creative. Exploring case studies of co-created art, music, and text, I argue that AI can be creative and demonstrate how AI can be creative and enhance our creativity. (shrink)

The AI duplication thesis claims that AI systems which are identical to humans on an abstract functional level of description relevant to psychology and behaviour (‘AI duplicates’) are nomologically possible. First, we argue that the AI duplication thesis should be understood in terms of the identity of medium-independent computational properties, and that medium-independence should not be understood as a special case of multiple realisability. Second, thus understood, AI duplicates are indeed nomologically possible. In favour of the first, we show that (...) the arguments against multiple realisability do not rule out that mental states are medium-independent computational states. In favour of the second, we show that the possibility of AI duplicates follows from the medium-independent character of computational properties, and the general ability to find equivalencies between medium-independent properties across systems. There are two key takeaways from our analysis. First, human minds are, to an important extent, independent of physical structure; very different kinds of materials, including of conventional AI, can function to implement their computational states. Second, we have removed one key obstacle to the view that artificial systems can have minds. If computational identity is sufficient for mental similarity/identity, then it would follow that AI systems can have minds. (shrink)

Some AI systems employ simulated representations of real-world scenarios, and these simulations are sometimes called ‘imagination’ in the engineering literature. Can imagination be attributed to AI systems? This paper will consider this question from a conceptual engineering perspective. Some key properties of imagination are identified, which characterise imagination’s relation to other mental states and its epistemic function. Simulations run by AI systems resemble human imagination with respect to these properties only partially. Thus a purely descriptive approach, which aims to address (...) the question of AI imagination by identifying the conditions that delineate our present concept of imagination, has limited prospects. Nevertheless, there are normative reasons to countenance AI imagination when simulation-based AI systems are seen as possible models of human intelligence. Some suggestions for adopting an AI-friendly concept of imagination will be offered. (shrink)

This paper explores the profound cognitive and affective impacts of Large Language Models (LLMs) and Generative AI (GenAI) on human thought, imagination, and especially our emotional life by analysing our interaction with different styles of GenAI agents especially when presented as virtual personalities. I describe how LLL-based agents can be presented using one of three broad metaphors of interaction, namely as Assistants, Extenders and Companions. Each metaphor of interaction implies distinctive cognitive patterns for their human interactant. I explore especially the (...) shape of some current interactions with GenAI companions and analyse their roles as cognitive and affective mind shapers. I analyse how Companion AI agents based on LLM technology can come to be treated as ‘Ersatz Others’ taking up deep roles in our cognitive and affective lives. I explore some of the perhaps unexpected implications of the deep reliance on such systems especially from the point of view of the theory of the extended mind and the mind-technology problem and show how this new sort of cognitive and affective environment for the human mind produces some new ethical challenges and may even reshape our future understandings of what minds are. (shrink)

Medical expertise is an essential concept when discussing the potential value of Artificial Intelligence (AI) systems in advancing the medical field, as one of AI’s main promises is to positively complement the expertise of medical professionals. Yet, relatively few articles on medical AI explicitly mention what medical expertise entails, how it is practiced in medical contexts, and how AI systems relate to medical professionals’ expertise. We believe it is essential to focus on answering these questions to determine how AI should (...) be integrated into medical practice and how we should formulate the expected benefits of AI systems in clinical care. With this article, we conceptualize medical expertise and identify four forms of medical expertise: epistemic expertise, performative expertise, adaptive expertise, and virtuous expertise. Additionally, we highlight how this conceptualization of expertise can support the ethical evaluation of AI’s augmenting role in medical practices. (shrink)

This paper investigates the role of Inference to the Best Explanation (IBE) in Explainable Artificial Intelligence (XAI), arguing that IBE can both act as a framework for revealing the considerations and preferences behind XAI explanations and for guiding and evaluating them in a more rigorous manner. The paper focusses specifically on post-hoc explainability and model agnostic techniques, illustrating the argument through examples of salience maps and feature importance techniques. Here, IBE in XAI happens in two steps. The first aims at (...) maximising the range and accuracy of possible inferences made; the second aims to provide a certain balance of explanatory virtues given a certain why-question and stakeholder that requires an explanation. Using the language of explanatory virtues enables a more nuanced discussion about what we may like or dislike about certain explanations in the current XAI landscape and provides actionable ways of improving explanations. (shrink)

Sensorimotor enactivism is the view that the content and the sensory modality of perceptual experience are determined by implicit knowledge of lawful regularities between bodily movements and patterns of sensory stimulation. A proponent of the explanatory argument for sensorimotor enactivism holds that this view is able to provide an intelligible explanation for why certain material realizers give rise to certain perceptual experiences, while rival accounts cannot close this “explanatory gap”. However, I argue that the notion of the “material realizer” of (...) perceptual experience is ambiguous. On a narrow construal, the explanatory gap cannot be bridged, not even by enactivism. On a wide construal, enactivism gets a grip on the explanatory gap, but to the same extent do established theories of consciousness. Thus, on both horns of the dilemma, the explanatory power of enactivism is not superior to more traditional theories of conscious experience. (shrink)

In the aftermath of the success of attention-based transformer networks, the debate over the potential and role of consciousness in artificial systems has intensified. Prominently, the Global Neuronal Workspace Theory emerges as a front-runner in the endeavor to model consciousness in computational terms. A recent advancement in the direction of mapping the theory onto state-of-the-art machine learning tools is the model of a Global Latent Workspace. It introduces a central latent representation around which multiple modules are constructed. Leveraging dedicated encoder-decoder (...) structures, content from the central representation or any individual module, integrated via the latent space, can be translated to any other module and back with minimal loss. This paper presents a thought experiment involving a minimal setup with one deep sensory and one deep motor module, which illustrates the emergence of “globally” accessible sensorimotor representations in the central latent space connecting both modules. In the human brain, neuronally enacted knowledge of laws relating changes in sensory information to changes in motor output or corresponding efferent copy information have been proposed to constitute the biological correlates of phenomenal conscious experience. The underlying Sensorimotor Contingency Theory encompasses a rich mathematical framework. Yet, the implementation of intelligent systems based on this framework has thus far been confined to proof-of-concept and basic prototype applications. Here, the natural appearance of global latent sensorimotor representations links two major neuroscientific theories of consciousness in a powerful machine learning setup. A remaining question is whether this artificial system is conscious. (shrink)

We describe a method, based on Jennifer Nado’s proposal for classification procedures as targets of conceptual engineering, that implements such procedures by prompting a large language model. We apply this method, using data from the Wikidata knowledge graph, to evaluate stipulative definitions related to two paradigmatic conceptual engineering projects: the International Astronomical Union’s redefinition of PLANET and Haslanger’s ameliorative analysis of WOMAN. Our results show that classification procedures built using our approach can exhibit good classification performance and, through the generation (...) of rationales for their classifications, can contribute to the identification of issues in either the definitions or the data against which they are being evaluated. We consider objections to this method, and discuss implications of this work for three aspects of theory and practice of conceptual engineering: the definition of its targets, empirical methods for their investigation, and their practical roles. The data and code used for our experiments, together with the experimental results, are available in a Github repository (https://github.com/bradleypallen/zero-shot-classifiers-for-conceptual-engineering). (shrink)