L. A. Paul

Drawing on philosophical theories of knowledge, I develop a conceptual framework for knowledge in animals and machines. I use this framework to engineer a new concept of large language model (LLM) knowledge and propose a taxonomy of knowledge that spans biological and artificial categories, from C. elegans to LLMs. The approach lays a foundation for a rigorous account of how knowledge can be realized across animals and machines while at the same time respecting the diverse biological and physical sources of such knowledge.

In certain problem-solving contexts, people organize their domain through treating themselves as the perceptual and cognitive center of their world. They identify and solve a particular problem from their perspective as a particular agent, with a particular location, at a particular time, in a particular environment. When they do this, they engage in a distinctive kind of agent-centered problem-solving. For many contexts involving intelligent agents, partially observable Markov decision processes (POMDPs), a framework for modeling decision-making in uncertain environments unfolding over time, have effectively become a “standard model.” Yet, as these models are ordinarily interpreted, they do not explicitly represent agent- centered problem-solving. Accordingly, to model this type of problem-solving, we begin by extending the standard POMDP framework to de ne “ePOMDPs.” This formalism models how an agent, once it centers itself on a particular self-and-world representation, plans and acts rationally from its own perspective. To capture the way that such agents choose which problem to solve, we build on our ePOMDPs to develop a“meta-ePOMDP” agent within a hierarchical Bayesian framework. We implement meta-ePOMDP agents for two different suites of“centering game” tasks that highlight different aspects of our theory. We nd that our models explain signatures of agent-centered problem-solving not captured by alternative models, in particular, the dif culty of navigating spaces of possible problem representations. We close by suggesting that our model could provide the beginnings of a computational framework for how a person could have a self.

Drawing together the metaphysics of counterfactuals with empirical work on intuitive judgments, this chapter discusses the nature of counterfactual reasoning about self-involving possibilities. It argues that when a person reasons about her self-involving possibilities, especially far-fetched possibilities, this reasoning may be supported by an underlying “self simulator,” a kind of mental engine with an approximate understanding of who she is, which enables her to learn about her preferences and make intuitive judgments and predictions about her self-involving possibilities. On this view, through observing or simulating their own choices, people understand themselves through a process similar to that by which they understand other people. In this way, they learn about their own beliefs and desires. The argument is informed by empirical data from surveys where lay people and philosophers decided what action they would take in vignettes involving a potentially transformative decision.

What has emerged from our discussion is that there is a deep divide in our intuitions about causation, and correspondingly, in how we want to handle two very central issues: problems with redundant causation, most notably with late preemption and overdetermination, and problems with causation involving omissions. We see a deep intuitive tension in the way we want to judge these cases, one that creates serious trouble for any reductive treatment that wants to approach causation in a uniform manner. And we are still looking for an acceptable, suitably reductive treatment of causation involving omissions, and a reductive way of combining causal structures while still drawing distinctions between the different types that are to be analyzed in different ways. We discuss how using the method by counterexample to search for a reductive theory of causation, and its sister search for an analysis of the concept of causation, generates an enormous bounty by opening up a number of interesting and potentially fruitful lines of enquiry, emphasize the importance of keeping track of differences in methodological approach, and close with the observation that a detailed understanding of the twists and turns in the discussion of causation is crucial for a range of debates in philosophy and debates in other fields, like economics, law, sociology and psychology.

Transitivity seems to underlie basic features of our causal reasoning: it is typical to justify a claim that _C_ causes _E_ by pointing out that _C_ causes _D_, which in turn causes _E_. Preserving transitivity seems to be a basic desideratum for an adequate analysis of causation, and appealing to it has often seemed essential for handling problem cases like those involving preemption. But recent work on causation has raised serious challenges to the claim that it is invariably transitive. The aim of this chapter is to distinguish these various challenges and to consider several responses to these challenges in detail. To that end, we critically discuss a causal modelling proposal and a contrastive proposal for handling challenging cases that combine transitivity with double prevention. We also discuss the nuances of switching cases, and show how to avoid mismatch of aspects by changing to a more subtle treatment of causes and effects.

This chapter focuses on examples that, in one way or another, involve omissions, that is, failures of events to occur. We discuss the metaphysical status of omissions, show how omission involving causation exhibits striking dissimilarities from ordinary causation, and look closely at structures involving causation by omission, prevention, and combinations of prevention and preemption. We then look carefully at the problems these structures create for reductive accounts of causation, including causal modeling accounts, and pay particular attention to some recent and sophisticated attempts to connect the treatment of omission-involving causation with a distinction between default states and deviant states. We conclude that we find ourselves in the position of being unable to dismiss causation involving omissions, unable to assimilate it to garden-variety causation, and unable to give a decently unified or complete account of it using any of the standard approaches to causation that we have canvassed.

This chapter introduces some of the problems for analyses of causation, sketches some of the most significant approaches to providing a philosophical account of causation, and discusses the methodological rules we intend to follow. We discuss counterfactual analyses, regularity-based accounts, causal modeling or structural equations accounts, contrastive accounts, de facto accounts, and transference accounts in more detail, with the aim of developing a cleaner understanding of what an account of causation ought to accomplish, and a better appreciation of the point of the close, examples-based kind of analysis featured in the next three chapters. We include a developed discussion of our methodological rules, we sketch some of the metaphysical assumptions we will make, and we discuss and set aside a number of issues that space precludes us from taking up in further detail.

Chapter 6 considers how peer disagreement over religion presents an epistemological problem: How can confidence in any religious claims including their negations be epistemically justified? Here, it is shown that the transformative nature of religious experience poses a further problem: to transition between religious belief and skepticism is not just to adopt a different set of beliefs, but to transform into a different version of oneself. It is argued that this intensifies the problem of pluralism by adding a new dimension to religious disagreement, for we can lack epistemic and affective access to our potential religious, agnostic, or skeptical selves. Yet, access to these selves seems to be required for the purposes of decision-making that is to be both rational and authentic. Finally, the chapter reflects on the relationship between the transformative problem and what it shows about the epistemic status of religious conversion and deconversion, in which one disagrees with one’s own transformed self.

Our paper contributes to the rapidly expanding body of experimental research on transformative decision making, and in the process, marks out a novel empirical interpretation for assessments of subjective value in transformative contexts. We start with a discussion of the role of subjective value in transformative decisions, and then critique extant experimental work that explores this role, with special attention to Reuter and Messerli (2018). We argue that current empirical treatments miss a crucial feature of practical deliberation manifesting across a variety of everyday decisions: often, people attach more weight to decision criteria that they can know—a phenomenon known as the “evaluability bias.” In transformative contexts, if people cannot know the subjective value of an option, they are unlikely to attach it much weight. Despite this, people may care very much about such value. We then use this point to develop and present new empirical results that, in addition to supporting our concerns about evaluability bias, support the hypothesis that people care about subjective value. Our work enriches the current philosophical understanding of transformative decisions and helps to frame the emerging experimental paradigm for the empirical dimensions of the debate.

I claim that Mill has a theory of poetry which he uses to reconcile nineteenth century associationist psychology, the tendency of the intellect to dissolve associations, and the need for educated members of society to desire utilitarian ends. The heart of the argument is that Mill thinks reading poetry encourages us to feel the feelings of others, and thus to develop pleasurable associations with the pleasurable feelings of others and painful associations with the painful feelings of others. Once the associations are developed, they are supported and maintained by our natural capacity for sympathy and by external elements in society, and provide motivation for the pursuit of utilitarian ends. Further, the additional support causes the associations to be strengthened to the extent that they come to be seen as ‘natural and necessary’, and as such are immune from the dissolving force of the intellect.

I address two related questions: first, what is the best theory of how objects have de re modal properties? Second, what is the best defence of essentialism given the variability of our modal intuitions? I critically discuss several theories of how objects have their de re modal properties and address the most threatening antiessentialist objection to essentialism: the variability of our modal intuitions. Drawing on linguistic treatments of vagueness and ambiguity, I show how essentialists can accommodate the variability of modal intuitions while holding that objects have their modal properties independently of contexts.

Causation is a deeply intuitive and familiar relation, gripped powerfully by common sense. Or so it seems. As is typical in philosophy, however, that deep intuitive familiarity has not led to any philosophical account of causation that is at once clean, precise, and widely agreed upon. Not for lack of trying: the last thirty years or so have seen dozens of attempts to provide such an account, and the pace of development is, if anything, accelerating. (See Collins et al. [2003a] for a comprehensive sampling of the latest work.)

A commitment to truth requires that you are open to receiving new evidence, even if that evidence contradicts your current beliefs. You should be open to changing your mind. However, this truism gives rise to the paradox of empathy. The paradox arises with the possibility of mental corruption through transformative change, and has consequences for how we should understand tolerance, disagreement, and the ability to have an open mind. I close with a discussion of how understanding this paradox provides a new explanation for a certain kind of standoff between the believer and the skeptic with regard to religious belief.

Among the many philosophers who hold that causal facts1 are to be explained in terms of—or more ambitiously, shown to reduce to—facts about what happens, together with facts about the fundamental laws that govern what happens, the clear favorite is an approach that sees counterfactual dependence as the key to such explanation or reduction. The paradigm examples of causation, so advocates of this approach tell us, are examples in which events c and e— the cause and its effect— both occur, but: had c not occurred, e would not have occurred either. From this starting point ideas proliferate in a vast profusion. But the remarkable disparity among these ideas should not obscure their common foundation. Neither should the diversity of opinion about the prospects for a philosophical analysis of causation obscure their importance. For even those philosophers who see these prospects as dim—perhaps because they suffer post-Quinean queasiness at the thought of any analysis of any concept of interest—can often be heard to say such things as that causal relations among events are somehow “a matter of” the patterns of counterfactual dependence to be found in them.

It seems natural to choose whether to have a child by reflecting on what it would be like to actually have a child. I argue that this natural approach fails. If you choose to become a parent, and your choice is based on projections about what you think it would be like for you to have a child, your choice is not rational. If you choose to remain childless, and your choice is based upon projections about what you think it would be like for you to have a child, your choice is not rational. This suggests we should reject our ordinary conception of how to make this life-changing decision, and raises general questions about how to rationally approach important life choices.