Desire, Decoupling, and Factive Mentalizing. Mind & Language (2026) [PhilPapers]
Why do we find it easier to attribute knowledge than belief? A promising answer suggests that attributing belief requires representations with content that can come apart from reality, whereas attributing knowledge does not. However, this account fails to generalize to desire attribution. Like belief, desire can diverge from the world, yet research suggests that attributing desire is fast, easy, and early developing—much more like attributing knowledge than belief. I propose an alternative explanation, the efficient recovery account, which explains differences between knowledge, belief, and desire attribution in terms of processing efficiency rather than representational content.
Reasons Isolated from Good Reasoning. Australasian Journal of Philosophy (2026) [PhilPapers]
McHugh and Way (2022) propose that reasons are premises in good reasoning, and that good defeasible reasoning is reasoning that manifests a pattern that normally preserves fittingness (i.e. normally, when the premises are fitting, the conclusion is fitting). On this proposal, rational defeat is explained by violations of normality (e.g. the defeater that John lied is abnormal). I argue this approach cannot handle two types of pro tanto reasons. Suppose someone offers you 35mg dose of heroin. That taking it is pleasurable is a reason to take it. Or suppose the first squirrel you see is grey. This observation is a reason to believe all squirrels are grey. Now suppose that there are no other reasons for taking the heroin or believing that all squirrels are grey---the reasons are isolated. I argue that McHugh and Way’s proposal cannot capture these cases of isolated normally overridden reasons or isolated weak reasons, even with significant modifications. Both only participate in bad patterns of reasoning. The pattern of reasoning that moves from the premise that the dose is pleasurable to the conclusion of an intention to take it fails to normally preserve fittingness. (Because the lethality of 35mg of heroin is normal, and thus intending to take it is not choiceworthy!) The pattern of reasoning that moves from the premise that this squirrel is grey to the conclusion that all squirrels are grey fails to normally preserve fittingness. (Because one observation is an insufficient inductive base!) These cases suggest that defeasibility isn’t explained by normality.
Implicit Beliefs and Other Peculiarities (with Ron Mallon). The Oxford Handbook of the Cognitive Science of Belief. Oxford University Press, 2027 [PhilPapers]
The past 40 years has seen an explosion of research using “implicit” or indirect measures, purportedly uncovering a host of implicit attitudes (IAs) that escape the net of traditional, direct measures. Researchers are especially interested in IAs that reflect social biases. We describe two influential methodologies of indirect measurement, including the Implicit Association Test (IAT), and then summarize important features of IAs. But what are IAs? What are indirect measures measuring? In this entry, we survey prominent accounts in psychology and philosophy of the nature of IAs including views on which IAs are associations, views on which they are beliefs, and challenges to these views that hold IAs are, instead, patchy endorsements, traits, imaginings, or mental imagery. We conclude by exploring the possibility of a pluralism on which IAs are a motley collection of real mental entities.
Treating As (manuscript available)---to be presented at the Eastern APA 2027
What is it to treat something as a reason? Philosophers are split on Cognitivism, the view that treating p as a reason requires representing p as a reason. This paper attempts to make progress on this debate by getting a better grip on talk of treating something as a reason. I do so by developing a general account of treating as ascriptions, using treating someone as a friend as a paradigm case. Drawing from research in formal semantics, I propose and defend a counterfactual and contextual semantics of these ascriptions. On this view, to treat o as an F is for your actual treatment of o to resemble your treatment of o in a normal world where you judge that o is an F, where the set of possible worlds is specified by the context. I argue this account is superior to several natural competitors and reveals new ways of understanding Cognitivism. The account indicates a path forward for understanding treating p as a reason: clarifying the nature of the context demanded by discussing rationality.
Naturalizing Inference (manuscript available)
Inference---conscious, deliberate, and person-level reasoning---presents a dilemma for the naturalist. Inference seems to transparently commit the reasoner to relations of support between premises and conclusions. Yet explaining how inference commits the reasoner in this way seems to require positing a fundamentally normative primitive, contra naturalism. In this paper, I propose how naturalists should understand inference. I develop the Reverse Taking Account of inference: the thinker draws her conclusion from her premises and takes her premises to support her conclusion because of this fact. I fill in the details of the account by drawing on accounts of inference from cognitive science. First, there are bare inferential transitions (BITs), transitions between discursive mental representations guided by rules built into the architecture of cognitive systems. In addition, there are rich inferential transitions (RITs), which are BITs that have caused, in the right way, the belief that the premises support the conclusion. Inferences are RITs on the Reverse Taking Account. I argue that the Reverse Taking Account of inference explains how inference can transmit internalist warrant and how it is personal-level and agential. RITs have the right causal-doxastic structure to secure the transmission of internalist warrant from premises to conclusions. RITs are firmly person-level and sufficiently agential to secure deontic status and responsibility assessment for inference. The naturalist can have their inferential cake and eat it too.
AI, Black-boxes, and Public Reason (in preparation)
AI systems are opaque. When these systems make decisions, we do not know what considerations are relevant to their decisions nor how those considerations relate to their decisions. Many have argued that this opacity is incompatible with the demands of public reason. Public reason requires that all moral and political decisions have reasons that can be publicly scrutinized, yet AI systems’ reasons cannot be scrutinized. However, this argument conflates distinct kinds of reasons: normative reasons, which are considerations in favor of some action or attitude, and motivating reasons, which are the reasons on the basis of which an agent acts/holds an attitude. Public reason, I argue, demands that all decisions have normative reasons that can be scrutinized, not necessarily motivating reasons. For example, it would not change the moral and legal legitimacy of Loving v. Virginia to discover that Chief Justice Warren, the author of the court’s opinion, was not motivated by the opinion but instead by some caprice. The court’s opinion remains a fully sufficient normative reason. Because contemporary reasoning models can produce fully sufficient normative reasons (e.g., a sound argument) for their decisions, at least some AI systems are compatible with the demands of public reason. That even reasoning models are opaque is no objection to their use within public reason. Surprisingly, in perhaps the most normatively significant domain, there is one less barrier to LLM use.
The Developmental Priority of Reasons (in preparation)
In this paper, I argue children deploy the reasons concept before other normative concepts (ought, good, etc.). Developmental psychology indicates that young children adroitly seek and accept answers to why-questions. I argue this competence can constitute the capacity to attribute reasons, and I identify markers for its possession. Young children meet these markers before there is evidence they deploy other normative concepts. This developmental priority suggests that reasons cannot be conceptually analyzed by other normative notions.
Kant on Sympathy and Affective Self-Governance (with Anne Margaret Baxley) (manuscript available)---to be presented at the Central APA 2027
Kant appears ambivalent about the moral value of sympathy. He is often wary of it as a source of moral motivation, yet he holds in the Doctrine of Virtue that the duty to sympathy [Teilnehmung] is a duty of virtue. Given this ambivalent stance, what is the moral function of sympathy and why is it morally valuable? Researchers have offered three broad answers to these questions: sympathy guides agents to morally important features (epistemic), allows agents to communicate moral importance (communicative), or can motivate right action (motivational). Despite the strengths of these accounts, we argue that each is inadequate. We propose a new account: sympathy is an affect regulator. Cultivating our capacities for sympathetic feeling and actively sympathizing can inhibit the negative affects that interfere with our autocracy, our ability to enforce and enact the dictates of reason. This account is broadly motivational, but sympathy does not directly motivate good action. Instead, it is a disabler of a disabler. The key idea is that sympathy’s moral function is in emotional regulation, helping imperfect rational agents inhibit the negative affects that undermine their freedom.
On Being Known (manuscript available)
Consider the person who has many intimates yet believes that others do not know her. Her belief could be both rational and true. Yet many accounts of interpersonal knowledge either don’t vindicate this belief or make knowing someone implausibly difficult. Acquaintance and De Te knowledge don’t vindicate the belief, as they are too easy for intimates to achieve. Verstehen/Perspective Taking in the humanistic tradition of understanding makes knowing someone implausibly difficult, as suggested by recent arguments for skepticism about empathy. I propose a view that escapes these concerns: to know someone is to know both her character traits and the personal history that explains them. This positivist proposal, rooted in the naturalist tradition of understanding, captures much of the important data yet makes knowing others challenging but possible. I argue it reveals a reason we want to be known: knowing these features enables others to influence us in ways that would otherwise be impossible, first observed by Plato in his account of psuchagogia [soul-leading].
Know-How Licenses Instruction/Knowledge-How Norm of Instruction (manuscripts available)---presented at Social Epistemology Network Event 4 @ Georgetown University
Both projects defend KHNI:
KHNI: One is epistemically properly positioned to instruct how to φ if and only if one knows how to φ.
"Know-How Licenses Instruction" defends the sufficiency direction, while "Knowledge-How Norm of Instruction" defends the necessity direction. Significant linguistic evidence converges upon KHNI, and the norm provides a strong rationalizing explanation of a litany of our practices. These are strong prima facie reasons for KHNI. Both papers are then concerned with dealing with counterexamples and disarming objections to KHNI. These papers advance our understanding of knowledge-how in three ways. First, they clarify the nature of instruction, allowing a clearer foundation for debates about the relation between knowledge-how and teaching. Second, they address prominent counterexamples to principles like KHNI that have not been seriously examined in the literature. Third, by showing the plausibility of KHNI, the papers reveal a new way to evaluate arguments about the nature of knowledge-how.