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)

Traditionally, perceptual experiences—for example, the experience of seeing a cat—were thought to have two quite distinct components. When one sees a cat, one’s experience is “about” the cat: this is the representational or intentional component of the experience. One’s experience also has phenomenal character: this is the sensational component of the experience. Although the intentional and sensational components at least typically go together, in principle they might come apart: the intentional component could be present without the sensational component or vice (...) versa. (shrink)

Under what conditions would an artificially intelligent system have wellbeing? Despite its clear bearing on the ethics of human interactions with artificial systems, this question has received little direct attention. Because all major theories of wellbeing hold that an individual’s welfare level is partially determined by their mental life, we begin by considering whether artificial systems have mental states. We show that a wide range of theories of mental states, when combined with leading theories of wellbeing, predict that certain existing (...) artificial systems have wellbeing. Along the way, we argue that there are good reasons to believe that artificial systems can have wellbeing even if they are not phenomenally conscious. While we do not claim to demonstrate conclusively that AI systems have wellbeing, we argue that there is a significant probability that some AI systems have or will soon have wellbeing, and that this should lead us to reassess our relationship with the intelligent systems we create. (shrink)

Claims about consciousness in animals are often made in support of their moral standing. Peter Carruthers argues that there is no fact of the matter about animal consciousness and it is of no scientific or ethical significance. Sympathy for an animal can be grounded in its mental states, but should not rely on assumptions about its consciousness.

Does the capacity to think require the capacity to sense? A lively debate on this topic runs throughout the history of philosophy and now animates discussions of artificial intelligence. Many have argued that AI systems such as large language models cannot think and understand if they lack sensory grounding. I argue that thought does not require sensory grounding: there can be pure thinkers who can think without any sensory capacities. As a result, the absence of sensory grounding does not entail (...) that large language models cannot think or understand. I also consider to what extent quasi-sensory grounding can at least boost the performance of a language model. (shrink)

According to hedonism about well-being, lives can go well or poorly for us just in virtue of our ability to feel pleasure and pain. Hedonism has had many advocates historically, but has relatively few nowadays. This is mainly due to three highly influential objections to it: The Philosophy of Swine, The Experience Machine, and The Resonance Constraint. In this paper, I attempt to revive hedonism. I begin by giving a precise new definition of it. I then argue that the right (...) motivation for it is the ‘experience requirement’ (i.e., that something can benefit or harm a being only if it affects the phenomenology of her experiences in some way). Next, I argue that hedonists should accept a felt-quality theory of pleasure, rather than an attitude-based theory. Finally, I offer new responses to the three objections. Central to my responses are (i) a distinction between experiencing a pleasure (i.e., having some pleasurable phenomenology) and being aware of that pleasure, and (ii) an emphasis on diversity in one’s pleasures. (shrink)

According to the experience requirement on well-being, differences in subjects’ levels of welfare or well-being require differences in the phenomenology of their experiences. I explain why the two existing arguments for this requirement are not successful. Then, I introduce a more promising argument for it: that unless we accept the requirement, we cannot plausibly explain why only sentient beings are welfare subjects. I argue, however, that because the right kind of theory of well-being can plausibly account for that apparent fact (...) about welfare subjects even if the requirement is false, this argument does not succeed. I tentatively conclude that no compelling case can be made for the requirement. (shrink)

Can we make progress exploring consciousness? Or is it forever beyond human reach? In science we never know the ultimate outcome of the journey. We can only take whatever steps our current knowledge affords. This paper explores today's evidence from the viewpoint of Global Workspace theory. First, we ask what kind of evidence has the most direct bearing on the question. The answer given here is ‘contrastive analysis’ -- a set of paired comparisons between similar conscious and unconscious processes. This (...) body of evidence is already quite large, and constrains any possible theory . Because it involves both conscious and unconscious events, it deals directly with our own subjective experience, as anyone can tell by trying the demonstrations in this article. (shrink)

What are the limits of well-being? This question nicely captures one of the central debates concerning the nature of the individual human good. For rival theories differ as to what sort of facts directly constitute a person's being well-off. On some views, well-being is limited to the presence of pleasure and the absence of pain. But other views push the boundaries of well-being beyond this, so that it encompasses a variety of mental states, not merely pleasure alone. Some theories then (...) draw the line here, limiting well-being to the presence of the appropriately broadened set of mental states. But still others extend the limits of well-being even further, so that it is constituted in part by facts that are not themselves mental states at all; on such views, well-being is partly constituted by states of affairs that are “external” to the individual's experiences. In this essay, I want to explore some of this debate by focusing on a particular stretch of the dialectic. That is, I want to think hard about a particular connected series of arguments and counterarguments. These arguments – or, at least, the concerns they seek to express – emerge naturally in the give and take of philosophical discussion. Together they make up a rather simple story, whose plot, in very rough terms, is this: first there is an attempt to push the limits of well-being outward, moving from a narrow to a broader conception; then comes the claim that the resulting notion is too broad, and so we must retreat to a narrower conception after all. (shrink)

Multiple realizability says that the same kind of mental states may be manifested by systems with very different physical constitutions. Putnam ( 1967 ) supposed it to be “overwhelmingly probable” that there exist psychological properties with different physical realizations in different creatures. But because function constrains possible physical realizers, this empirical bet is far less favorable than it might initially have seemed, especially when we take on board the richer picture of neural and brain function that neuroscience has been uncovering (...) over the past forty years, in which all sorts of brain activities play crucial roles beyond all-or-nothing electrical signaling. Because of its evolutionary history, the brain’s metabolic and informational processes are inextricably intertwined. The resulting complex integrated functions impose more constraints on possible realizers than the clean, single-purpose functions usually cited as examples in discussions of multiple realizability—with ramifications for the functionalist and computationalist foundations of cognitive science, artificial intelligence, and the philosophy of mind. (shrink)

One way to express the most persistent part of the mind-body problem is to say that there is an “explanatory gap” between the physical and the mental. The gap is not usually taken to apply to all of the mental, but to subjective experience, the mind’s “qualitative” features, or what is now referred to as “phenomenal consciousness.” The “gap” formulation is due to Joseph Levine. He acknowledged the appeal of intuitions of separability between physical facts, of any kind we can (...) envisage, and this aspect of our mental lives. Subjective experience seems not to be the sort of thing that is just the physical under another guise. The immediate focus of Levine’s discussion was a family of arguments due to Saul Kripke. Kripke argued that some influential claims of mind-body identity could not be, as materialists claimed, contingent. If these identities are real then they are necessary. But we can clearly conceive that the mental and physical could come apart—an intuition that “identity theorists” conceded. Given that the identity is necessarily present if present at all, from the fact it is not necessary we can see it is absent. Kripke concluded that physicalism is false. Levine wanted to resist this. (shrink)

The issue of the biological origin of consciousness is linked to that of its function. One source of evidence in this regard is the contrast between the types of information that are and are not included within its compass. Consciousness presents us with a stable arena for our actions—the world—but excludes awareness of the multiple sensory and sensorimotor transformations through which the image of that world is extracted from the confounding influence of self-produced motion of multiple receptor arrays mounted on (...) multijointed and swivelling body parts. Likewise excluded are the complex orchestrations of thousands of muscle movements routinely involved in the pursuit of our goals. This suggests that consciousness arose as a solution to problems in the logistics of decision making in mobile animals with centralized brains, and has correspondingly ancient roots. (shrink)

This paper proposes a brain-inspired cognitive architecture that incorporates approximations to the concepts of consciousness, imagination, and emotion. To emulate the empirically established cognitive efficacy of conscious as opposed to non-conscious information processing in the mammalian brain, the architecture adopts a model of information flow from global workspace theory. Cognitive functions such as anticipation and planning are realised through internal simulation of interaction with the environment. Action selection, in both actual and internally simulated interaction with the environment, is mediated by (...) affect. An implementation of the architecture is described which is based on weightless neurons and is used to control a simulated robot. (shrink)

Baars (1988, 1997) has proposed a psychological theory of consciousness, called global workspace theory. The present study describes a software agent implementation of that theory, called ''Conscious'' Mattie (CMattie). CMattie operates in a clerical domain from within a UNIX operating system, sending messages and interpreting messages in natural language that organize seminars at a university. CMattie fleshes out global workspace theory with a detailed computational model that integrates contemporary architectures in cognitive science and artificial intelligence. Baars (1997) lists the psychological (...) ''facts that any complete theory of consciousness must explain'' in his appendix to In the Theater of Consciousness; global workspace theory was designed to explain these ''facts.'' The present article discusses how the design of CMattie accounts for these facts and thereby the extent to which it implements global workspace theory. (shrink)

The last decade has experienced a vivid enthusiasm to unravel the mystery of consciousness believed to be one of the major puzzles of human kind. We share this enthusiasm. Still, we feel that current models are incomplete suffering from a problem that we call the “small network argument”.

After discussing various types of consciousness, several approaches to machine consciousness, software agent, and global workspace theory, we describe a software agent, IDA, that is 'conscious' in the sense of implementing that theory of consciousness. IDA perceives, remembers, deliberates, negotiates, and selects actions, sometimes 'consciously'. She uses a variety of mechanisms, each of which is briefly described. It's tempting to think of her as a conscious artifact. Is such a view in any way justified? The remainder of the paper considers (...) this question. (shrink)