Jonathan Birch

Some patients, following brain injury, do not outwardly respond to spoken commands, yet show patterns of brain activity that indicate responsiveness. This is ‘cognitive-motor dissociation’ (CMD). Recent research has used machine learning to diagnose CMD from electroencephalogram recordings. These techniques have high false discovery rates, raising a serious problem of inductive risk. It is no solution to communicate the false discovery rates directly to the patient’s family, because this information may confuse, alarm and mislead. Instead, we need a procedure for generating case-specific probabilistic assessments that can be communicated clearly. This article constructs a possible procedure with three key elements: (1) A shift from categorical ‘responding or not’ assessments to degrees of evidence; (2) The use of patient-centred priors to convert degrees of evidence to probabilistic assessments; and (3) The use of standardised probability yardsticks to convey those assessments as clearly as possible.

نواجه تحدّيين عاجلين فيما يتعلّق بالوعي والذكاء الاصطناعي. يتمثّل التحدّي الأول في أنّ ملايين المستخدمين سيُسيئون قريبًا إسناد وعيٍ شبيهٍ بالوعي البشري إلى أصدقاء الذكاء الاصطناعي وشركائهم ومساعديهم، استنادًا إلى المحاكاة ولعب الأدوار، ولا نعرف كيف يمكن منع ذلك. أمّا التحدّي الثاني فهو أنّ أشكالًا غريبةً وعميقةً للغاية من الوعي قد تتحقق فعلًا في الذكاء الاصطناعي، لكن فهمنا النظري للوعي لا يزال غير ناضج بما يكفي لتقديم إجابات واثقة، سواء بالإيجاب أو بالنفي. والموقف الوسطي بشأن وعي الذكاء الاصطناعي هو أنّ علينا أخذ هذين التحدّيَين على محمل الجد. ويتفاعل التحدّيان بطرق تجعل الأمر صعبًا؛ فالإجراءات الرامية إلى معالجة التحدّي الأول قد تُضعف محاولات معالجة التحدّي الثاني عبر تصوير فكرة الوعي الاصطناعي على أنّها مستحيلة أو غير مرجّحة بطبيعتها. وفي المقابل، قد تؤدي محاولات معالجة التحدّي الثاني إلى ارتفاع مستويات إسناد الوعي الخاطئ لدى المستخدمين العاديين. ويحاول هذا "البيان" صياغة استراتيجيات متسقة متبادلة للتعامل مع كلا التحدّيين.

Nick Bostrom’s ‘Simulation Argument’ purports to show that, unless we are confident that advanced ‘posthuman’ civilizations are either extremely rare or extremely rarely interested in running simulations of their own ancestors, we should assign significant credence to the hypothesis that we are simulated. I argue that Bostrom does not succeed in grounding this constraint on credence. I first show that the Simulation Argument requires a curious form of selective scepticism, for it presupposes that we possess good evidence for claims about the physical limits of computation and yet lack good evidence for claims about our own physical constitution. I then show that two ways of modifying the argument so as to remove the need for this presupposition fail to preserve the original conclusion. Finally, I argue that, while there are unusual circumstances in which Bostrom’s selective scepticism might be reasonable, we do not currently find ourselves in such circumstances. There is no good reason to uphold the selective scepticism the Simulation Argument presupposes. There is thus no good reason to believe its conclusion.

Rapid advances in artificial intelligence (AI) have led users to believe that systems such as large language models (LLMs) have mental states, including the capacity for ‘experience’ (e.g., emotions and consciousness). These folk-psychological attributions often diverge from expert opinion and are distinct from attributions of ‘intelligence’ (e.g., reasoning, planning), and yet may affect trust in AI systems. While past work provides some support for a link between anthropomorphism and trust, the impact of attributions of consciousness and other aspects of mentality on user trust remains unclear. We explored this in a preregistered experiment (N = 410) in which participants rated the capacity of an LLM to exhibit consciousness and a variety of other mental states. They then completed a decision-making task where they could revise their choices based on the advice of an LLM. Bayesian analyses revealed strong evidence against a positive correlation between attributions of consciousness and advice-taking; indeed, a dimension of mental states related to experience showed a negative relationship with advice-taking, while attributions of intelligence were strongly correlated with advice acceptance. These findings highlight how users’ attitudes and behaviours are shaped by sophisticated intuitions about the capacities of LLMs—with different aspects of mental state attribution predicting people’s trust in these systems.

I contrast my picture of the relationship between the science and policy of animal sentience with that of Marian Stamp Dawkins, who thinks “the science of animal sentience and the politics of animal welfare should be kept separate” because they involve irreconcilably different standards of evidence. On my alternative picture, (i) the science of animal sentience, like any other empirical science, delivers evidence but not certainty; (ii) this evidence allows us to make better practical decisions, both within and outside science and (iii) the quality standards we apply to the evidence should be high in all contexts, including the formulation of public policy.

Definitions of animal welfare often invoke consciousness or sentience. Marian Stamp Dawkins has argued that to define animal welfare this way is a mistake. In Dawkins’s alternative view, an animal with good welfare is one that is healthy and “has what it wants.” The dispute highlights a source of strain on the concept of animal welfare: consciousness-involving definitions are better able to capture the normative significance of welfare, whereas consciousness-free definitions facilitate the validation of welfare indicators. I reflect on how the field should respond to this strain, ultimately recommending against splitting the concept and in favor of consciousness-involving definitions.

I briefly present and motivate a 'skill hypothesis' regarding the evolution of human normative cognition. On this hypothesis, the capacity to internally represent action-guiding norms evolved as a solution to the distinctive problems of standardizing, learning and teaching complex motor skills and craft skills, especially skills related to toolmaking. We have an evolved cognitive architecture for internalizing norms of technique, which was then co-opted for a rich array of social functions. There was a gradual expansion of the normative domain, with ritual playing an important role in bridging the gap between concrete, enacted norms and general, abstract norms, such as kinship norms. I conclude by stating nine predictions arising from the skill hypothesis.

This is a response to the nine commentaries on our target article “Unlimited Associative Learning: A primer and some predictions”. Our responses are organized by theme rather than by author. We present a minimal functional architecture for Unlimited Associative Learning (UAL) that aims to tie to together the list of capacities presented in the target article. We explain why we discount higher-order thought (HOT) theories of consciousness. We respond to the criticism that we have overplayed the importance of learning and underplayed the importance of spatial modelling. We decline the invitation to add a negative marker to our proposed positive marker so as to rule out consciousness in plants, but we nonetheless maintain that there is no positive evidence of consciousness in plants. We close by discussing how UAL relates to development and to its material substrate.

Kantian ethics has struggled terribly with the challenge of incorporating non-human animals as beings to which we can owe obligations. Christine Korsgaard’s Fellow Creatures is a bold, substantial attempt to meet that challenge. It is a significant book. In this essay, I won’t attempt a chapter-by-chapter summary. I will set the scene for the book’s core argument (which comes rather late, in Chapter 8), offer a reconstruction of that argument, and reflect on its strengths and limitations. I’ll also briefly compare the view at which Korsgaard arrives to that of Tom Regan in The Case for Animal Rights (1983).

Simona Ginsburg and Eva Jablonka: The Evolution of the Sensitive Soul: Learning and the Origins of Consciousness. MIT Press, Cambridge, MA, 2019, 646pp, ISBN: 9780262039307 The Evolution of the Sensitive Soul is a landmark attempt to make progress on the problem of animal consciousness. Ginsburg and Jablonka propose a general cognitive marker of the presence of consciousness: Unlimited Associative Learning. They use this marker to defend a generous view about the distribution of consciousness in the natural world, on which a capacity for conscious experience is common to all vertebrates, many arthropods and some cephalopod molluscs. They use this inferred distribution to defend a view about the evolution of consciousness, on which it has evolved at least three times, first evolved at around the time of the Cambrian explosion (just over 500 million years ago), and was in fact the driving force behind that explosion. In this essay review, I reflect critically on the book’s central idea: the proposal that Unlimited Associative Learning provides a general marker of consciousness.

This paper attempts to reconcile critics and defenders of inclusive fitness by constructing a synthesis that does justice to the insights of both. I argue that criticisms of the regression-based version of Hamilton’s rule, although they undermine its use for predictive purposes, do not undermine its use as an organizing framework for social evolution research. I argue that the assumptions underlying the concept of inclusive fitness, conceived as a causal property of an individual organism, are unlikely to be exactly true in real populations, but they are approximately true given a specific type of weak selection that Hamilton took, on independent grounds, to be responsible for the cumulative assembly of complex adaptation. Finally, I reflect on the uses and limitations of “design thinking” in social evolution research. The debate about the foundations of inclusive fitness theory that has followed in the wake of Nowak, Tarnita and Wilson’s (1) critique has been remarkably polarizing. After several rounds of rebuttals and replies, there is still little evidence of any serious reconciliation between the theory’s critics (2-10) and its defenders (11-24). It doesn’t have to be this way. I believe that, on the main points of disagreement, it is possible to find a way forward that does justice to the insights of both camps. My aim in this paper is to find that way forward.

Hamilton's theory of kin selection is the best-known framework for understanding the evolution of social behavior but has long been a source of controversy in evolutionary biology. A recent critique of the theory by Nowak, Tarnita, and Wilson sparked a new round of debate, which shows no signs of abating. In this overview, we highlight a number of conceptual issues that lie at the heart of the current debate. We begin by emphasizing that there are various alternative formulations of Hamilton's rule, including a general version, which is always true; an approximate version, which assumes weak selection; and a special version, which demands other restrictive assumptions. We then examine the relationship between the neighbor-modulated fitness and inclusive fitness approaches to kin selection. Finally, we consider the often-strained relationship between the theories of kin and multilevel selection.

We consider the relationship between neural and behavioural evidence for animal consciousness. We critically examine two recent studies: one neural and one behavioural. The first, on crows, finds different neural activity depending on whether a stimulus is reported as seen or unseen. However, to implicate this neural activity in consciousness, we must assume that a specific conditioned behaviour is a report of conscious experience. The second study, on macaques, records behaviours strikingly similar to patterns of conscious and unconscious perception in humans. However, confounds are only ruled out in human subjects, presupposing substantial neural similarity between humans and macaques. Taken together, the two studies reveal a sense in which neural and behavioural research rely on each other. Looking ahead, these two types of evidence could prove to be either mutually reinforcing or mutually undermining. The science of animal consciousness needs both neural and behavioural evidence, ideally obtained as part of a single coordinated programme.

Consciousness has an important role in ethics: when a being consciously experiences the frustration or satisfaction of its interests, those interests deserve higher moral priority than those of a behaviourally similar but non-conscious being. I consider the relationship between this ethical role and an a posteriori (or ‘type-B’) materialist solution to the mind-body problem. It is hard to avoid the conclusion that, if type-B materialism is correct, then the reference of the concept Phenomenal Consciousness is radically indeterminate between a neuronal-level property that is distinctive to mammals and a high-level functional property that is much more widely shared. This would leave many non-mammalian animals (such as birds, fish, insects and octopuses) with indeterminate moral status. There are ways to manage this radical moral indeterminacy, but all of these ways lead to profoundly troubling consequences.

I reflect on the commentaries on my 'skill hypothesis' from Andrews/Westra, Tomasello, Sterelny, and Railton. I discuss the difference between normative cognition and the broader category of action-guiding representation, and I reflect on the relationship between joint intentionality and normative cognition. I then consider Sterelny and Railton's variants on the skill hypothesis, which highlight some important areas where future evidence could help us refine the account: The relative importance of on-The-fly skill execution vs. longer-Term strategizing, the relative importance of toolmaking vs. collaborative foraging, and the question of whether norms are encoded in control models themselves or in the goals and ideals that our control models help us pursue.

What makes fast, cumulative cultural evolution work? Where did it come from? Why is it the sole preserve of humans? We set out a self-assembly hypothesis: cultural evolution evolved culturally. We present an evolutionary account that shows this hypothesis to be coherent, plausible, and worthy of further investigation. It has the following steps: (0) in common with other animals, early hominins had significant capacity for social learning; (1) knowledge and skills learned by offspring from their parents began to spread because bearers had more offspring, a process we call CS1 (or Cultural Selection 1); (2) CS1 shaped attentional learning biases; (3) these attentional biases were augmented by explicit learning biases (judgements about what should be copied from whom). Explicit learning biases enabled (4) the high-fidelity, exclusive copying required for fast cultural accumulation of knowledge and skills by a process we call CS2 (or Cultural Selection 2) and (5) the emergence of cognitive processes such as imitation, mindreading and metacognition-'cognitive gadgets' specialized for cultural learning. This self-assembly hypothesis is consistent with archaeological evidence that the stone tools used by early hominins were not dependent on fast, cumulative cultural evolution, and suggests new priorities for research on 'animal culture'. This article is part of the theme issue 'Foundations of cultural evolution'.

Over the past two decades, Ginsburg and Jablonka have developed a novel approach to studying the evolutionary origins of consciousness: the Unlimited Associative Learning (UAL) framework. The central idea is that there is a distinctive type of learning that can serve as a transition marker for the evolutionary transition from non-conscious to conscious life. The goal of this paper is to stimulate discussion of the framework by providing a primer on its key claims (Part I) and a clear statement of its main empirical predictions (Part II).

There is increasing recognition that the welfare needs of cephalopod molluscs and decapod crustaceans are important. Current commercial practices involving these animals include a range of potential threats to their welfare, such as conditions of farming, capture, transport, and slaughter. This article draws from and updates our 2021 review for the UK Government, recommending a range of relatively simple and impactful changes that could benefit welfare while highlighting important research gaps that should be prioritised to facilitate the drafting of guidelines for best-practice.

There is no agreement on whether any invertebrates are conscious and no agreement on a methodology that could settle the issue. How can the debate move forward? I distinguish three broad types of approach: theory-heavy, theory-neutral and theory-light. Theory-heavy and theory-neutral approaches face serious problems, motivating a middle path: the theory-light approach. At the core of the theory-light approach is a minimal commitment about the relation between phenomenal consciousness and cognition that is compatible with many specific theories of consciousness: the hypothesis that phenomenally conscious perception of a stimulus facilitates, relative to unconscious perception, a cluster of cognitive abilities in relation to that stimulus. This “facilitation hypothesis” can productively guide inquiry into invertebrate consciousness. What is needed? At this stage, not more theory, and not more undirected data gathering. What is needed is a systematic search for consciousness-linked cognitive abilities, their relationships to each other, and their sensitivity to masking.

Skyrms, building on the work of Dretske, has recently developed a novel information-theoretic account of propositional content in simple signalling systems. Information-theoretic accounts of content traditionally struggle to accommodate the possibility of misrepresentation, and I show that Skyrms’s account is no exception. I proceed to argue, however, that a modified version of Skyrms’s account can overcome this problem. On my proposed account, the propositional content of a signal is determined not by the information that it actually carries, but by the information that it would carry at the nearest separating equilibrium of the underlying evolutionary dynamics. I show that this amended account yields reasonable ascriptions of false propositional content in a well-known formal model of the evolution of communication (the ‘Philip Sidney’ game), and close with a discussion of the serious but perhaps not insuperable difficulties we face in applying the account to examples of signalling in the real world.

The theories of inclusive fitness and multilevel selection provide alternative perspectives on social evolution. The question of whether these perspectives are of equal generality remains a divisive issue. In an analysis based on the Price equation, Queller argued (by means of a principle he called the separation condition) that the two approaches are subject to the same limitations, arising from their fundamentally quantitative-genetical character. Recently, van Veelen et al. have challenged Queller’s results, using this as the basis for a broader critique of the Price equation, the separation condition, and the very notion of inclusive fitness. Here we show that the van Veelen et al. model, when analyzed in the way Queller intended, confirms rather than refutes his original conclusions. We thereby confirm (i) that Queller’s separation condition remains a legitimate theoretical principle and (ii) that the standard inclusive fitness and multilevel approaches are indeed subject to the same limitations.