Derek Shiller

The Moral Parliament Tool provides a framework for modeling participants' general moral views and simulating democratic deliberation among them. We show how the tool can be used to support democratic decisions about how to allocate healthcare resources. We apply the tool to three different bioethics case studies, showing how a hypothetical moral parliament composed of delegates of diverse worldviews would decide to distribute resources. This approach provides a more generalizable alternative to existing methods for participatory algorithm design.

We develop a taxonomical framework for classifying challenges to the possibility of consciousness in digital artificial intelligence systems. This framework allows us to identify the level of granularity at which a given challenge is intended (the levels we propose correspond to Marr's levels) and to disambiguate its degree of force: is it a challenge to computational functionalism that leaves the possibility of digital consciousness open (degree 1), a practical challenge to digital consciousness that suggests improbability without claiming impossibility (degree 2), or an argument claiming that digital consciousness is strictly impossible (degree 3)? We apply this framework to 14 prominent examples from the scientific and philosophical literature. Our aim is not to take a side in the debate, but to provide structure and a tool for disambiguating between challenges to computational functionalism and challenges to digital consciousness, as well as between different ways of parsing such challenges.

Some of the technical details of the implementation of modern AI systems raise novel questions about computational functionalism. In particular, GPU clusters processing large language models often interweave the computations for many different chat responses. If the computations underlying these language models were to give rise to consciousness when run in isolation, it is not clear what to make of their status when they are interwoven. This paper draws on debates in the personal identity literature to present two alternative options, one in which many minds can be interwoven in one GPU cluster, and one on which at most a single mind can exist. Each position is coherent and it is not clear how to resolve the central issues. This is likely to be a source of further uncertainty about consciousness in production AI setups.

Functionalist theories of consciousness attribute consciousness to a system when it displays specific patterns of behaviors and state changes in response to perceptual events and its prior state. These patterns can be disrupted in ways that would intuitively not preclude consciousness. The challenge for functionalists is to explain why pattern-disrupting influences do not remove consciousness. This paper explores this issue in connection with a simple case. Bombs may make it unlikely that nearby brains will manifest typical behaviors. Non-exploding bombs do not preclude consciousness within systems in their vicinity no matter how likely they are to go off. The challenge is trickier than it may first appear: this paper surveys possible responses and raises issues for each.

The prospect of consciousness in artificial systems is closely tied to the viability of functionalism about consciousness. Even if consciousness arises from the abstract functional relationships between the parts of a system, it does not follow that any digital system that implements the right functional organization would be conscious. Functionalism requires constraints on what it takes to properly implement an organization. Existing proposals for constraints on implementation relate to the integrity of the parts and states of the realizers of roles in a functional organization. This paper presents and motivates three novel integrity constraints on proper implementation not satisfied by current neural network models. It is proposed that for a system to be conscious, there must be a straightforward relationship between the material entities that compose the system and the realizers of functional roles, that the realizers of the functional roles must play their roles due to internal causal powers, and that they must continue to exist over time.

ABSTRACT If the future is highly sensitive to the past, then many of our acts have long-term consequences whose significance well exceeds that of their foreseeable short-term consequences. According to an influential argument by James Lenman, we should think that the future is highly sensitive to acts that affect people’s identities. However, given the assumption that chancy events are ubiquitous, the effects that our acts have are likely to dissipate over a short span of time. The sets of possible futures left open by alternative acts are typically very similar in the same way that large random samples drawn from the same population are typically very similar.

Jaegwon Kim has developed an argument that interactionist dualists cannot account for the causal relations between minds and brains. This paper develops a closely related argument that focuses instead on the causal relations between minds and neurons. While there are several promising responses to Kim’s argument, their plausibility relies on a relatively simple understanding of mind–brain relations. Once we shift our focus to neurons, these responses lose their appeal. The problem is that even if mind–brain causal pairing can be explained at no great theoretical cost, the complex interactions non-physical minds would need to have with neurons cannot.

This paper offers a noncognitivist characterization of moral attitudes, according to which moral attitudes count as such because of their inclusion of moral concepts. Moral concepts are distinguished by their contribution to the functional roles of some of the attitudes in which they can occur. They have no particular functional role in other attitudes, and should instead be viewed as evolutionary spandrels. In order to make the counter-intuitive implications of the view more palatable, the paper ends with an account of the evolution of normative judgments as exaptations of the cognitive structures that underlie beliefs.

Non-cognitivists are known to face a problem in extending their account of straightforward predicative moral judgments to logically complex moral judgments. This paper presents a related problem concerning how non-cognitivists might extend their accounts of moral judgments to other kinds of moral attitudes, such as moral hopes and moral intuitions. Non-cognitivists must solve three separate challenges: they must explain the natures of these other attitudes, they must explain why they count as moral attitudes, and they must explain why the moral attitudes are systematically correlated with ordinary propositional attitudes. After presenting the problem, this paper examines several contemporary theories with some initial promise for solving it, and argues that they are insufficient.

In this paper, I propose that those who reject higher-order theories of consciousness should not rule out the possibility of having conscious experiences that they cannot introspect. I begin by offering four arguments that such non-introspectible conscious experiences are possible. Next, I offer two arguments for thinking that we actually have such experiences. According to the first argument, it is unlikely that evolution would have furnished us with a faculty of introspection that worked flawlessly. According to the second argument, there are many plausible potential sources of non-introspectible experiences. Given that all of these sources are at least somewhat plausible, it is fairly probable that we have some kind or other. Finally, I consider whether we might be justified in believing that we can introspect all of our conscious experiences if in fact we can. I show that current approaches to justifying belief in epiphenomenal qualia do not carry over. The upshot of these arguments is that we should be less certain that we really know what it feels like to be us. There may be much more to our experiences than we are aware of.

If it is within our power to provide a significantly better world for future generations at a comparatively small cost to ourselves, we have a strong moral reason to do so. One way of providing a significantly better world may involve replacing our species with something better. It is plausible that in the not‐too‐distant future, we will be able to create artificially intelligent creatures with whatever physical and psychological traits we choose. Granted this assumption, it is argued that we should engineer our extinction so that our planet's resources can be devoted to making artificial creatures with better lives.

It has recently been alleged that expressivism cannot account for the obvious fact that normative sentences and their negations express inconsistent kinds of attitudes. I explain how the expressivist can respond to this objection. I offer an account of attitudinal inconsistency that takes it to be a combination of descriptive and normative relations. The account I offer to explain these relations relies on a combination of functionalism about normative judgments and expressivism about the norms governing them. It holds that the inconsistency of normative judgments is primitive. One potential problem for this view is that the large number of normative primitives that the expressivist will allegedly need to accept will render the view grossly unparsimonious. In defending this thesis, I suggest that it is a mistake to hold the lack of normative parsimony of expressivism against its core psychological claims.