I am a philosopher working in ethics and epistemology, with interests in the philosophy of AI and applied ethics.
I recently completed my PhD in Philosophy at the University of Texas at Austin, where I was supervised by John Bengson and Jonathan Dancy. Before my PhD, I studied English at Oxford, theology at Harvard, and obtained an MA in philosophy from Georgia State University. I am originally from Ebbw Vale, in the South Wales Valleys.
I have taught philosophy to 562 students across 17 classes, serving as primary instructor for 14. See Teaching below for more details.
Papers
-
Autonomous Weapons: Proportionality, Probability, and UncertaintyUnder review.
Suppose that, when deploying an autonomous weapon system (AWS), we have no justified expectation that the resulting non-combatant harm will be excessive, but we also have no justified expectation that it will not be excessive. Is deploying the AWS in this case permissible? This paper argues, first, that AWS deployments will often meet these conditions and, second, that such cases expose a neglected problem in the ethics of proportionality under uncertainty.
Firstly, I argue that, in many cases of AWS deployment, it will not be feasible to calculate the probability of excessive harm that will result. Moreover, the standard method of constructing an “assurance case” for weapons safety will not generally provide us with a justified expectation that excessive non-combatant harm will not result from the decision to deploy AWS.
Secondly, I distinguish between cases in which the military advantage anticipated from deployment is low and those in which it is high. In the former cases, both interpretations of proportionality considered here entail that it is only permissible to deploy AWS when we limit the circumstances of deployment such that the maximum plausible non-combatant harm is not excessive. However, when the anticipated advantage is high, whether the principle permits deployment depends on controversial questions about the kind of justified expectation that proportionality requires of us. More philosophical work is needed on how uncertain threats to innocent bystanders should be weighed against harms that defenders justifiably expect if they refrain from acting.
-
Black Box Testimony: When Are We Justified in Believing LLM Outputs?Under review.
It is hard to know how to assess the outputs of large language models (LLMs). One common response to this debate is what I will call the Dismissive View, i.e. the view that we are justified in believing LLM outputs only if we corroborate them with a reputable, independent source. This paper argues both that the Dismissive View is false and that nevertheless LLM outputs do not by default justify belief. I instead argue that by default they justify a weaker epistemic attitude that I call “considering.” In arguing this, I draw on the epistemology of testimony without assuming that LLM outputs are themselves a form of testimony. In particular, I appeal to Tyler Burge’s non-reductionist theory of testimony in order to explain why LLM outputs possess a kind of prima facie credibility in virtue of their connection to human rationality. Burge’s theory helps explain why we are justified in believing certain kinds of LLM output without further corroboration, in particular those reporting easily verifiable facts about which the training data is likely to be in consensus. However, Burge’s theory fails to distinguish between the various ways in which intelligible expressions might relate to rationality, and the different kinds of justificatory force they may have as a result. I argue that, by default, we are prima facie justified in considering the propositional contents of LLM outputs: keeping them in mind as something whose truth or implications merit some further attention, without thereby believing or inquiring into them.
-
On Hunches and IntuitionsUnder review.
Sometimes a proposition stands out to us even when its truth does not seem obvious and we do not seem to have reason to believe it. This experience is the kind of mental state I call a ``hunch.'' This paper argues that hunches are a distinctive kind of mental state that can play an important epistemic role, particularly under conditions of deep uncertainty. Hunches are not reducible to familiar accounts of belief, guessing, or seemings. Moreover, they are not captured by the prominent perceptualist account of intuitions. I argue that they are conscious, non-voluntary mental suggestions that make it seem fitting to consider their contents. Although a hunch need not justify believing its content, it can justify considering it. In this way, hunches can guide attention and deliberation under conditions of uncertainty, helping to shape what we eventually believe, understand, and do. If hunches are a kind of intuition then we need a more inclusive account of what intuitions are than the perceptualist account. If hunches are not intuitions, then we ought to pay more attention to the informative mental states that exist outside this category.
-
Did Socrates Want to be Executed? The Rhetoric of Pleasure and Pain in the ApologyRevise and resubmit at an ancient philosophy journal.
Interpretations of Socrates’ conduct in Plato’s Apology have tended to favour one of two views: that he deliberately provoked the jury into executing him, or that he sincerely tried to secure acquittal. Both interpretations explain the form of his defence primarily in terms of his attitude toward his own survival. I argue, however, that neither of these two accounts of Socrates’ behaviour at trial is as persuasive as the one suggested by Socrates himself in the Gorgias. There, Socrates claims that if someone were ever to bring him to trial unjustly, he would not attempt to persuade the judges to acquit him. Instead, he suggests that his speeches would be devoted to making his fellow citizens more just rather than to pleading his own case. On my interpretation, Socrates’ speeches in the Apology are calculated to expose the injustices of his audience and to begin the process of punishing those injustices so as to improve the souls of his hearers. Socrates was therefore neither deliberately attempting to secure his own execution nor attempting to persuade the judges to acquit him in the interests of justice. His famously provocative speeches are instead primarily intended to make his hearers more just.
Presentations
Teaching
Student evaluations
Very Good or Excellent±
92.5%
Of 117 respondents at UT Austin, 92.5% rated the classes I designed as either “Very Good” or “Excellent,” with 54% selecting “Excellent.”
Inclusive learning environment±
98.4%
98.4% of students surveyed at UT Austin agreed or strongly agreed that I fostered an inclusive learning environment; the remaining 1.6% selected “Neutral.”
Applied ethical reasoning±
100%
Students surveyed in my UT Austin ethics classes unanimously either agreed or strongly agreed that the assignments improved their ability to apply ethical reasoning skills in real-life situations.
Selected student comments
“It was almost as if we were trying to solve a problem together and looking at things from different perspectives to get a better angle at the issue at hand.”Fall 2023
“He made people's ideas really feel valued, not by complimenting and applauding them, but by taking them seriously and earnestly engaging with them.”Spring 2020
“I felt more comfortable presenting my opinions as the semester went on knowing that you would encourage me to critically engage with and elaborate on my thoughts. This being a philosophy class full of opinionated men, I never thought that would be possible.”Spring 2020