There is an apparent connection between reinforcement learning and agency. Artificial entities controlled by reinforcement learning algorithms are standardly referred to as agents, and the mainstream view in the psychology and neuroscience of agency is that humans and other animals are reinforcement learners. This article examines this connection, focusing on artificial reinforcement learning systems and assuming that there are various forms of agency. Artificial reinforcement learning systems satisfy plausible conditions for minimal agency, and those which use models of the environment (...) to perform forward search are capable of a form of agency which may reasonably be called action for reasons. (shrink)

In some recent work on omissions, it has been argued that the causal theory of action cannot account for how agency is exercised in intentionally omitting to act in the same way it explains how agency is exercised in intentional action. Thus, causalism appears to provide us with an incomplete picture of intentional agency. I argue that causalists should distinguish causalism as a general theory of intentional agency from causalism as a theory of intentional action. Specifically, I argue that, while (...) intentional actions may best be understood as the causal products or outcomes of causings, we should identify exercises of intentional agency with causal processes. With a causalist account of intentional agency sketched, I respond to the challenge to causalism from omissions. I argue that when an agent intentionally omits there is a causal process that has a zero-sum outcome. But the causal process is sufficient to make it true that the agent exercises intentional agency in intentionally omitting. (shrink)

No philosopher has worked harder than Dan Dennett to set the possibility of machine mentality on firm philosophical footing. Dennett’s defense of this possibility has both a positive and a negative thrust. On the positive side, he has developed an account of mental activity that is tailor-made for the attribution of intentional states to purely mechanical contrivances, while on the negative side, he pillories as mystery mongering and skyhook grasping any attempts to erect barriers to the conception of machine mentality (...) by excavating gulfs to keep us “bona fide” thinkers apart from the rest of creation. While I think he’s “won” the rhetorical tilts with his philosophical adversaries, I worry that Dennett’s negative side sometimes gets the better of him, and that this obscures advances that can be made on the positive side of his program. In this paper, I show that Dennett is much too dismissive of original intentionality in particular, and that this notion can be put to good theoretical use after all. Though deployed to distinguish different grades of mentality, it can (and should) be incorporated into a philosophical account of the mind that is recognizably Dennettian in spirit. (shrink)

The ability to use and make technical artifacts has been considered exclusive to human beings. However, recent findings in ethology in light of observations made in nature and in laboratory show the opposite. In the area of philosophy of technology there are few exceptions that take into account the ability of some non-human animals to manufacture and use tools. In this paper I want to show some reasons to reconsider other possibilities. It seems that capacities such as intentionality, culture or (...) even the complexity of the structures of manufactured object are not exclusive to human beings. I will suggest a different way of analyzing objects created by non-human animals, one that tries to explain the gradualness of the structure, but also the plasticity of the behavior exhibited by non-human animals. These two elements allow a deeper understanding of the great variety of objects that other animals also manufacture and use. (shrink)

Recently, many philosophers have been inclined to ascribe mentality to animals on the main grounds that they possess certain complex computational abilities. In this paper I contend that this view is misleading, since it wrongly assumes that those computational abilities demand a psychological explanation. On the contrary, they can be just characterised from a computational level of explanation, which picks up a domain of computation and information processing that is common to many computing systems but is autonomous from the domain (...) of psychology. Thus, I propose that it is possible to conceive insects and other animals as mere computing agents, without having any commitment to ascribe mentality to them. I conclude by sketching a proposal about how to draw the line between mere computing and genuine mentality. (shrink)

Fred Dretske’s account of mental causation, developed in Explaining Behavior and defended in numerous articles, is generally regarded as one of the most interesting and most ambitious approaches in the field. According to Dretske, meaning facts, construed historically as facts about the indicator functions of internal states, are the structuring causes of behavior. In this article, we argue that Dretske’s view is untenable: On closer examination, the real structuring causes of behavior turn out to be markedly different from Dretske’s meaning (...) facts. Our argument proceeds in three steps. First, we set forth the problem of meaning individuation: We argue that the proposal that meaning facts are structuring causes of behavior commits Dretske to a very fine-grained individuation of meanings that is deeply counterintuitive. In a second step, we show that even these finely individuated meaning facts cannot do the job that they are supposed to do, since information facts—which are constitutive of, but distinct from Dretske’s meaning facts—are better candidates for the role of structuring causes. Finally, we argue that it is not even information facts, but facts of co-instantiation which are the real structuring causes of behavior. In concluding, we briefly consider the options that are left for Dretske if our arguments succeed. (shrink)

Often, the behavior of animals can be better explained and predicted, it seems, if we ascribe the capacity to have beliefs, intentions, and concepts to them. Whether we really can do so, however, is a debated issue. Particularly, Donald Davidson maintains that there is no basis in fact for ascribing propositional attitudes or concepts to animals. I will consider his and rival views, such as Colin Allen's three-part approach, for determining whether animals possess concepts. To avoid pure theoretical debate, however, (...) I will test these criteria using characteristic examples from ethology that depict a broad range of animal behavior. This will allow us to detect a series of gradations in animals' capacities, in the course of which we can think over what would count for or against an attribution of concepts and propositional attitudes to them in each single case. Self-conceit is our natural hereditary disease. Of all creatures man is the most wretched and fragile, and at once the most supercilious. ... It is by this conceit that man arrogates to himself ... divine properties, that he segregates himself from the mass of other creatures and raises himself above them .. (shrink)

We discuss the evolutionary transition from animals with limited experiencing to animals with unlimited experiencing and basic consciousness. This transition was, we suggest, intimately linked with the evolution of associative learning and with flexible reward systems based on, and modifiable by, learning. During associative learning, new pathways relating stimuli and effects are formed within a highly integrated and continuously active nervous system. We argue that the memory traces left by such new stimulus-effect relations form dynamic, flexible, and varied global sensory (...) states, which we call categorizing sensory states . These CSSs acquired a function: they came to act as internal “evaluators” and led to positive and negative reinforcement of new behavior. They are therefore the simplest, distinct, first-person motivational states that an animal can have. They constitute what we call basic consciousness, and are the hallmark of animals that can experience. Since associative learning has been found in many invertebrate taxa that first appeared during the Cambrian era, we propose that the processes underlying basic consciousness are phylogenetically ancient, and that their emergence may have fueled the Cambrian explosion. (shrink)

Current Perspectives in Philosophy of Biology.

The Chilean Cybersyn project, an attempt to manage a nation’s economy by cybernetic methods, has evoked more and more interest in recent years. The project’s design lead and several team members were alumni of the Ulm School of Design—an institution that has been labelled “Bauhaus successor” and today is famous for a no-arts and method-led design approach with strong societal aspirations. The school also influenced the emerging design discipline in Latin America during the 1960s and 70s. This article reviews topics (...) in the Ulm curriculum that influenced “Ulmian” thinking, but often remained unnoticed in design centred publications. Cybernetics, Operations Research and Information Theory and their relation to design are discussed and the related scholars are portrayed critically. (shrink)

Organisms and other biological entities are mistake-prone: they get things wrong. The entities of pure physics, such as atoms and inorganic molecules, do not make mistakes: they do what they do according to physical law, with no room for error except on the part of the physicist or their theory. We set out a novel framework for understanding biology and its demarcation from physics – that of mistake-making. We distinguish biological mistakes from mere failures. We then propose a rigorous definition (...) of mistakes that, although invoking the concept of function, is compatible with various views about what functions are. The definition of mistake-making is agential, since mistakes do not just happen ¬– at least in the sense analysed here – but are made. This requires, then, a notion of biological agency which we set out as a definition of the Minimal Biological Agent. The paper then considers a series of objections to the theory presented here, along with our replies. Two key features of our theory of mistakes are, first, that it is a supplement to, not a replacement for, existing general frameworks within which biology is understood and practised. Secondly, it is designed to be experimentally productive. Hence we end with a series of case studies where mistake theory can be shown to be useful in the potential generation of research questions and novel hypotheses of interest to the working biologist. (shrink)

In the recent literature on concepts, two extreme positions concerning animal minds are predominant: the one that animals possess neither concepts nor beliefs, and the one that some animals possess concepts as well as beliefs. A characteristic feature of this controversy is the lack of consensus on the criteria for possessing a concept or having a belief. Addressing this deficit, we propose a new theory of concepts which takes recent case studies of complex animal behavior into account. The main aim (...) of the paper is to present an epistemic theory of concepts and to defend a detailed theory of criteria for having concepts. The distinction between nonconceptual, conceptual, and propositional representations is inherent to this theory. Accordingly, it can be reasonably argued that some animals, e.g., grey parrots and apes, operate on conceptual representations. (shrink)

Morgan's canon can be construed as claiming that an intentional explanation of a behavior should be ruled out if there exists an explanation of this behavior in terms of 'lower' mechanisms. Unfortunately, Morgan's conception of higher and lower faculties is based on dubious evolutionary considerations. I examine alternative interpretations of the terms 'higher' and 'lower', and show that none can turn the canon into a principle that is both correct and useful in drawing the line between thinkers and non-thinkers. In (...) the process, I identify a number of problems that an adequate formulation of the canon should avoid. I then consider two more recent versions of the canon, proposed by Elliott Sober and Jonathan Bennett. Both are found unsatisfactory, but I argue that a version of Bennett's unity condition that is restricted to the attribution of recognitional concepts is on the right track. (shrink)

The aim of this exploratory paper is to review an under-appreciated parallel between group agency and artificial intelligence. As both phenomena involve non-human goal-directed agents that can make a difference to the social world, they raise some similar moral and regulatory challenges, which require us to rethink some of our anthropocentric moral assumptions. Are humans always responsible for those entities’ actions, or could the entities bear responsibility themselves? Could the entities engage in normative reasoning? Could they even have rights and (...) a moral status? I will tentatively defend the (increasingly widely held) view that, under certain conditions, artificial intelligent systems, like corporate entities, might qualify as responsible moral agents and as holders of limited rights and legal personhood. I will further suggest that regulators should permit the use of autonomous artificial systems in high-stakes settings only if they are engineered to function as moral (not just intentional) agents and/or there is some liability-transfer arrangement in place. I will finally raise the possibility that if artificial systems ever became phenomenally conscious, there might be a case for extending a stronger moral status to them, but argue that, as of now, this remains very hypothetical. (shrink)

There is a consensus among philosophers that some “I”-thoughts are immune to error through misidentification. In some recent papers, this property has been formulated in the following deflationist way: an “I”-thought is immune to error through misidentification when it can misrepresent the mental or bodily property self-ascribed but cannot misrepresent the subject possessing that property. However, it has been put forward that the range of mental and bodily states that are immune in that limited sense cannot include nonconceptual forms of (...) self-representation. In this paper, I claim the opposite. I argue in favor of a theoretical framework inspired by semantic relativism that solves the problem of immune nonconceptual self-representations. In order to do so, I refute an argument against the relativist account which is based on the existence of shared representations. This argument, I contend, rests on a confusion between two conditions to which a relativist may appeal when considering whether a certain mental content is relative to the self: a strong invariance condition and a weak invariance condition. I then argue that even if we acknowledge the existence of shared representations, the weak invariance condition is still satisfied, and consequently the relativist framework can make sense of INSRs. I argue that this weak invariance condition is satisfied by a representational function that self-relativizes certain representations. I then provide an empirical instance of such a function by discussing some of the recent literature on motor representations and the sense of agency. In the last part of the paper, I answer several potential objections. These potential objections lead me to distinguish two fundamental kinds of error relative to the self: error through misidentification and error through misapplication. This distinction allows me to answer a fundamental question raised by the very idea of de facto immunity to error through misidentification. (shrink)

Five mental components of human intentionality are distinguished and related to different properties of mammalian orientation. It is proposed that, in the course of evolution, these old properties became integrated and thereby allowed for the development of a new quality: human orientation. The existence of more than 4,000 mammal species with their various forms and levels of mental organization, offering a panorama of different combinations of differently developed components of mentality, provide ample opportunities for comparative studies. The difficulties in assessing (...) specific types are outlined, drawing on over 40 years of observation. Based on this knowledge, an argument is made for the importance of staying in contact with the empirical objects and of considering their ontological status when rising the standards of precision of formal analysis. (shrink)

Herder already very early in his career, in the 1760s, established two vitally important and epoch-making principles in the philosophy of language: that thought is essentially dependent on and bounded by language; and that meanings or concepts should be identified - not with such items as the referents involved, Platonic forms, or empiricist 'ideas' - but with word-usages. What did Herder do for an encore? His Treatise on the Origin of Language from 1772 might seem the natural place to look (...) for an answer to this question (since it is his best known work in the philosophy of language by far), but it is really the wrong place to look, because it temporarily regresses to a more conventional and less philosophically interesting position. However, Herder did succeed in making impressive progress in a broader array of works, namely by striving to identify prima facie problem cases confronting his two principles and to reconcile them with the latter. The main ones which he identified were God, animals, and non-linguistic art. In each of these cases, having initially proposed a reconciliation which did not work, he went on to develop a much more plausible one, indeed one which (at least in the two cases that really require one: animals and non-linguistic art) seems broadly correct. (shrink)

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)

Dignity is traditionally thought to apply to almost all and almost only humans. However, I argue that an account of a distinctly human dignity cannot achieve a coherent and non-arbitrary justification; either it must exclude some humans or include some nonhumans. This conclusion is not as worrying as might be first thought. Rather than attempting to vindicate human dignity, dignity should extend beyond the human, to include a range of nonhuman animals. Not only can we develop a widely inclusive account (...) of dignity by pursuing this route, but we can still defend the three core principles that lie at the heart of contemporary thinking about dignity: that bearers of dignity possess dignity to an equal degree, in virtue of possessing the same intrinsic worth, and that this generates direct and claimable rights. I thus develop an account of dignity that includes nonhuman animals. I argue that the capacity to value most plausibly grounds dignity. This capacity arises from valenced sentience: a being values insofar as they have experiences that matter to them. I then characterise what follows from this. First, I contend that it is only the capacity to value grounds dignity because this produces the most plausible, simplest, and most explanatorily powerful account of dignity. We ought to reject pluralist accounts which assert a sharp distinction between bearers and non-bearers of rational agency, and biocentric accounts which include non-sentient organic entities. Second, I argue this property grounds an account of dignity that is non-hierarchical. There are no degrees of dignity. Instead, all possessors of the capacity to value have the same kind of intrinsic value, and the same equal fundamental right: to consideration of their interest in leading a life of value. Finally, I argue that since the fundamental right is to consideration, moral agents have claimable duties to engage in a deliberative procedure with actionable requirements. Specific rights are instantiations of this general procedure. Certain revisionist conclusions follow on my view, for instance, that nonhumans should be given greater weight and inclusion in a far wider range of cases, but my account is sensitive to the range of different things different beings value, so it does not generate counterintuitive implications. I conclude that dignity extends beyond the human. (shrink)

If a person's head moves, she may or may not have moved her head, and, if she did move it, she may have actively performed the movement of her head or merely, by doing something else, caused a passive movement. And, if she performed the movement, she might have done so intentionally or not. This short array of contrasts (and others like them) has motivated questions about the nature, variety, and identity of action. Beyond the matter of her moving, when (...) the person moves her head, she may be indicating agreement or shaking an insect off her ear. Should we think of the consequences, conventional or causal, of physical behavior as constituents of an action distinct from but ‘generated by’ the movement? Or should we think that there is a single action describable in a host of ways? Also, actions, in even the most minimal sense, seem to be essentially ‘active’. But, how can we explain what this property amounts to and defend our wavering intuitions about which events fall in the category of the ‘active’ and which do not? (shrink)

I develop an interdisciplinary framework for understanding the nature of agents and agency that is compatible with recent developments in the metaphysics of science and that also does justice to the mechanistic and normative characteristics of agents and agency as they are understood in moral philosophy, social psychology, neuroscience, robotics, and economics. The framework I develop is internal perspectivalist. That is to say, it counts agents as real in a perspective-dependent way, but not in a way that depends on an (...) external perspective. Whether or not something counts as an agent depends on whether it is able to have a certain kind of perspective. My approach differs from many others by treating possession of a perspective as more basic than the possession of agency, representational content/vehicles, cognition, intentions, goals, concepts, or mental or psychological states; these latter capabilities require the former, not the other way around. I explain what it means for a system to be able to have a perspective at all, beginning with simple cases in biology, and show how self-contained normative perspectives about proper function and control can emerge from mechanisms with relatively simple dynamics. I then describe how increasingly complex control architectures can become organized that allow for more complex perspectives that approach agency. Next, I provide my own account of the kind of perspective that is necessary for agency itself, the goal being to provide a reference against which other accounts can be compared. Finally, I introduce a crucial distinction that is necessary for understanding human agency: that between inclinational and committal agency, and venture a hypothesis about how the normative perspective underlying committal agency might be mechanistically realized. (shrink)

My dissertation addresses the question "do desires provide reasons?" I present two independent lines of argument in support of the conclusion that they do not. The first line of argument emerges from the way I circumscribe the concept of a desire. Complications aside, I conceive of a desire as a member of a family of attitudes that have imperative content, understood as content that displays doability-conditions rather than truth-conditions. Moreover, I hold that an attitude may provide reasons only if it (...) has truth-evaluable content. Insofar as desires lack truth-evaluable content, I hold that the content of a desire has the wrong kind of logical structure to provide reasons. My second line of argument claims that even if a desire did have truth-evaluable content, it would not follow that desires provide reasons. This is because a desire has no more rational significance than a guess or coin-flip. My argument relies on what I call the non-substitutability principle, the thesis that (all things being equal) one cannot substitute something that lacks rational significance, relative to some attitude, A, for something that has rational significance, relative to A, and leave the rational standing of A unchanged. For example, one cannot substitute the guess that P (i.e., something that lacks rational significance relative to the belief that P) for the perception that P (i.e., something that is rationally significant relative to the belief that P) without altering the rational standing of the belief. I argue that when the non-substitutability principle is applied to a desire that gives rise to an intention, it turns out that one can always substitute a guess or coin-flip (i.e., something that lacks rational significance relative to the intention) for the desire, without altering the rational standing of the intention. I take this to show that desires are not rationally significant relative to the intentions to which they give. (shrink)