Research

Publications

• "A Qualified Defense of Top-down Approaches in Machine Ethics." AI & Society. Forthcoming.

Abstract:

This paper concerns top-down approaches in machine ethics. It is divided into three main parts. First, I briefly describe top-down design approaches, and in doing so I make clear what those approaches are committed to and what they involve when it comes to training an AI to behave ethically. In the second part, I formulate two underappreciated motivations for endorsing them, one relating to predictability of machine behavior and the other relating to scrutability of machine decision-making. Finally, I present three major worries about such approaches, and I attempt to show that advocates of top- down approaches have some plausible avenues of response. I focus most of my attention on what I call the ‘technical manual objection’ to top-down approaches, inspired by the work of Annas (2004). In short, the idea is that top-down approaches treat ethical decision-making as being merely a matter of following some ethical instructions in the same way that one might follow some set of instructions contained in a technical manual (e.g. computer manual), and this invites sensible skepticism about the ethical wisdom of machines that have been trained on those approaches. I respond by claiming that the objection is successful only if it is understood as targeting machines that have certain kinds of goals, and it should not compel us to totally abandon top-down approaches. Such approaches could still be reasonably employed to design ethical AI that operate in contexts that include fairly noncontroversial answers to ethical questions. In fact, we should prefer top-down approaches when it comes to those types of context, or so I argue, due to the advantages I claim for them.

• "Robust Artificial Moral Agents and Metanormativity." Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society. 2023.

Abstract:

This paper explores the relationship between our ignorance concerning certain metanormative topics and the design of ethical artificial intelligence (AI). In particular, it will be maintained that because we cannot predict in advance which metanormative conclusions a sufficiently intelligent ethical AI might reach, we have reason to be apprehensive about the project of designing such AI. Even if we succeeded at designing an AI to engage in ethical behavior, there is a distinct possibility that the AI might eventually cease to behave ethically if it reaches certain metanormative conclusions. The candidate conclusions include ones such as the denial of the alleged authority or overridingness of ethics and the conclusion that there are no ethical facts or properties (i.e. moral error theory). It will be argued that the target AI could conceivably reach such conclusions, and in turn this could cause them to abandon their ethical routines and proceed to cause great harm.

• "Deontologists Can Be Moderate." The Journal of Value Inquiry 52.2 (2018): 199-212.

Abstract:

Moderate deontology, the view that deontological constraints can be permissibly violated when and only when doing so prevents the occurrence of sufficiently bad consequences, has become a popular alternative to absolutist forms of deontology, which hold that deontological constraints can never be permissibly violated. It is a view that many find plausible because it accommodates commonsense deontological constraints, but it also permits commonsense violations of those constraints whenever very much is at stake (e.g. it permits one murder whenever committing a murder would prevent one million comparable murders). Considering the abundance of moderate deontologists, one would suspect that moderate deontology is probably a coherent, deontological position. However, with respect to its being a deontological position at all, Saul Smilansky maintains that the view is actually pluralist, not deontological, and that we should understand deontology only in its typical absolutist form. The objective of this essay is to show that, contra Smilansky, moderate deontology is properly understood as a deontological theory, and I hope to accomplish some conceptual clarification in the process regarding certain aspects of the theory. In particular, I will emphasize the primacy of deontological constraints in moderate deontology, discuss the normative implications of permissible constraints violations, and conclude with a succinct explanation of a point on which I partly agree with Smilansky concerning the significance of terminology in our normative theorizing.

Works in progress

• "A Case for End-constrained Ethical Artificial Intelligence" (under review)

Abstract:

Some people who advocate for the design of ethical AI argue that because AI could potentially harm us, we should focus some of our efforts in AI research on the design of ethical machines. Some authors have attempted to raise a problem for the foregoing line of thought by claiming that we do not need ethical AI in order to prevent harm to humans. We simply need safe machines. In this paper, I consider one such argument and raise some objections to it. In particular, I argue that the goal of implementing safety features in AI systems is in itself a much more complicated task than some authors seem to acknowledge, and moreover I maintain that merely safe AI could still be ethically problematic in numerous ways. Then, I show that a certain kind of ethical AI, which I call end-autonomous ethical AI, would be especially dangerous. I characterize them as having a strong form of autonomy (akin to humans) that allows for the possibility of numerous risks. Finally, I motivate the case for a specific category of ethical machines, namely, end-constrained ethical AI. I describe these AI as possessing whatever capacities would be necessary for satisfying ethical aims beyond safety while lacking end-autonomy. I argue that we should want to design end-constrained ethical AI for reasons extending beyond those of safety, and I emphasize some desirable capabilities they could have as well as some risks. I also mention a few examples of possible contexts into which such AI could be introduced and expected to behave ethically. In short, the aim of this paper is to establish that end-constrained ethical AI occupy a desirable middle ground between merely safe AI and end-autonomous ethical AI because they allow us to secure more ethical goods than just safety, but they are also not as ethically risky as end-autonomous ethical AI.

• "Reductive Ethical Pluralism and the Intelligibility of Moral Concern" (under review)

Abstract:

This paper concerns the relationship between two sorts of theses: metaethical ones regarding the reducibility of moral properties and first-order normative ones regarding how many different things are, say, good or right-making features of actions. It will be argued that multiple realizability of moral properties invites a distinctive challenge for ethical reductionists, which centers on the intelligibility of moral concern. The general idea is that an adequate metaethical theory of moral properties should be able to make sense of our ordinary moral concerns, and this would include both coherence with our pre-theoretical conceptions of moral properties and endowment of such properties with genuine explanatory power, and to the extent that the reduction bases posited by ethical reductionists for certain moral properties are disjunctive (or pluralistic), this makes things more difficult for the reductionist. I begin with some preliminary setup, which involves defining some relevant terminology and briefly describing some of the existing literature on the debate between ethical reductionists and non-reductionists that is pertinent to the aims of this essay. Next, I explain and argue for what I take to be a desideratum for metaethical theories of moral properties, and I maintain that the desideratum presents a unique challenge to reductive metaethical theories that are also pluralist. Finally, I show how two alternative realist theories satisfy the desideratum and conclude that ethical reductionists should prefer monism at the first-order level of theorizing or otherwise abandon reductionism in light of the arguments of the paper.

• "Decision Procedures for Artificial Moral Agents" (under review)

Abstract:

This paper explores the possibility that the appropriate decision procedures for artificial moral agents (AMAs) to utilize in their ethical decision-making are importantly different from the ones that are appropriate for human moral agents. It argues that the appropriate type of decision procedure for a given moral agent depends on the nature of the agent’s capacities, and thus certain kinds of AMAs should employ different decision procedures than the ones humans should use. If this conclusion is correct, then it has significant consequences for a number of issues, including the design of ethical artificial intelligence, the paradox of hedonism (and related puzzles), and the concept of virtue as it relates to AMAs. These consequences are discussed, and it is concluded that our commonsense views about certain ethical topics should be reconsidered in light of the relevant differences between artificial and human moral agents.

• "Conscious Artifical Moral Agents Mitigate Existential Risk"

Abstract:

[in progress]