We draw a distinction between the traditional reference class problem which describes an obstruction to estimating a single individual probability—which we re-term the individual reference class problem—and what we call the reference class problem at scale, which can result when using tools from statistics and machine learning to systematically make predictions about many individual probabilities simultaneously. We argue that scale actually helps to mitigate the reference class problem, and purely statistical tools can be used to efficiently minimize the reference class problem at scale, even though they cannot be used to solve the individual reference class problem.

We propose an approach to the evolution of joint agency and cooperative behavior that contrasts with views that take joint agency to be a uniquely human trait. We argue that there is huge variation in cooperative behavior and that while much human cooperative behavior may be explained by invoking cognitively rich capacities, there is cooperative behavior that does not require such explanation. On both comparative and theoretical grounds, complex cognition is not necessary for forms of joint action, or the evolution of cooperation. As a result, promising evolutionary approaches to cooperative behavior should explain how it arises across many contexts.

The definition of tool use has long been debated, especially beyond humans. Recent work argues that the phenomena included within tool use are so broad and varied that there is little hope of using the category for scientific generalizations, explanations, and predictions about the evolution, ecology, and psychology of tool users. One response to this argument has been the development of tooling as a replacement for tool use. In this article, we analyze the tool use and tooling frameworks. Identifying advantages and limitations in each, we offer a synthetic approach that suggests promising avenues for future research.

Douglas has argued that if values were accorded a ‘direct’ role during the ‘internal’ phases of science, this would amount to ‘wishful thinking’ in place of evidence. This paper draws on two claims that jointly threaten this position. First, building a cogent ethical case in favour of a claim about what ought to happen is not a simple matter of saying what one wishes were true; a good ethical case has a kind of discipline to it. Second, some theorists have argued that scientists do and should defend ‘mixed hypotheses’; that is, hypotheses that implicate both factual and evaluative content.

As technology fosters connections among like-minded individuals, concerns about the effects of homogeneous clusters—often criticized as ideological bubbles and echo chambers—have intensified. While these clusters are commonly seen as obstacles to independent thought and progress, this paper argues that they can, under certain conditions, drive significant advancements. By revising computational models of collective problem solving and examining historical cases, I demonstrate that clusters, particularly among minority groups with superior ideas, can overcome dominant resistance and promote progress. However, this clustering introduces trade-offs, including slower consensus formation.

The frequency of major theory change in natural science is rapidly decreasing. Sprenger and Hartmann (2019) claim that this observation can improve the justificatory basis of scientific realism, by way of what can be called a stability argument. By enriching the conceptual basis of Sprenger and Hartmann’s argument, this paper shows that stability arguments pose a strong and novel challenge to scientific anti-realists. However, an anti-realist response to this challenge is also proposed. The resulting dialectic establishes a level of meaningful disagreement about the significance of stability arguments for scientific realism, and indicates how the disagreement can ultimately be resolved.

It is sometimes assumed that Large Language Models (LLMs) know language, or for example that they know that Paris is the capital of France. But what—if anything—do LLMs actually know? In this paper, I argue that LLMs can acquire tacit knowledge as defined by Martin Davies (1990). Whereas Davies himself denies that neural networks can acquire tacit knowledge, I demonstrate that certain architectural features of LLMs satisfy the constraints of semantic description, syntactic structure, and causal systematicity. Thus, tacit knowledge may serve as a conceptual framework for describing, explaining, and intervening on LLMs and their behavior.

Three popular objections to reductionism about color, experience as of impossible colors, the unary/binary distinction, and structural mismatch, are issues just, I argue, for the (probably default) version of reductionism according to which colors reduce to sets of surface spectral reflectances (SSRs). They are not problems for the version on which colors are dispositions to reflect coarse-grained intensities of light—are what in colorimetry are called “object colors”. This paper sets out to demonstrate the virtues of the latter reductionism.

We present an agent-based model to study how the network structure of a scientific community could impact the public uptake of science, and how this impact is influenced by scientific uncertainty and affinity bias. For unbiased agents, a highly connected scientific network decreases the probability that the public favors the correct theory. For biased agents, however, a moderately connected scientific network causes the public to favor the correct theory more often. This results from the competition between the scarcity of information (for poorly connected agents) and the spread of misleading information (for highly connected agents). Adding more scientists strengthens both effects.

‘Relativistic thermodynamics’ should be understood not as a generalization of a non-relativistic theory but as an application of a general thermodynamic framework, neutral as to spacetime setting and allowing arbitrary conserved quantities, to the specific case of relativity. That framework gives an unambiguous result as to the thermodynamics of relativistically moving systems (an answer coinciding with Einstein’s, and Planck’s, original results.) Thermodynamic temperature is unambiguously defined as rate of change of energy with entropy at constant momentum; that said, its operational significance is limited and other measures of energy/entropy covariance, which incorporate momentum transfer, are often more useful.

In this paper I present the immunological account of physiological individuality courtesy of Thomas Pradeu (2012) and the evolutionary account of biological individuality from Ellen Clarke (2012, 2013). I argue that in combination, the logic of these two accounts implies that all physiological individuals are capable of undergoing evolution by natural selection. The main objection to this view is the case of holobionts (Godfrey-Smith 2013, Pradeu 2016). Here, I will argue that this objection is unjustified and that holobionts meet basic criteria for evolutionary individuality. As such, this supports the view that physiological individuals are also evolutionary individuals.

There is an “under-representation problem” in philosophy departments and journals. Empirical data suggest that while we have seen some improvements since the 1990s, the rate of change has slowed down. Some posit that philosophy has disciplinary norms making it uniquely resistant to change. We present results from an empirical case study of a philosophy department that achieved and maintained male-female gender parity among its faculty as early as 2014. Our analysis extends beyond matters of gender parity because that is only one, albeit important, dimension of inclusion. We build from the study to reflect on strategies that may catalyze change.

Scientific fields frequently need to exchange data to advance their own inquiries. Data unification is the process of stabilizing these forms of interfield data exchange. I present an account of the epistemic structure of data unification, drawing on case studies from model-based cognitive neuroscience (MBCN). MBCN is distinctive because it shows that modeling practices play an essential role in mediating these data exchanges. Models often serve as interfield evidential integrators, and models built for this purpose have their own representational and inferential functions. This form of data unification should be seen as autonomous from other forms, particularly explanatory unification.

We evaluate the roles general relativistic assumptions play in simulations used in recent observations of black holes including LIGO-Virgo and the Event Horizon Telescope. In both experiments simulations play an ampliative role, enabling the extraction of more information from the data than would be possible otherwise. This comes at a cost of theory-ladenness. We discuss the issue of inferential circularity, which arises in some applications; classify some of the epistemic strategies used to reduce the extent of theory-ladenness; and discuss ways in which these strategies are model independent.

Does low mood have intentional content? If so, what is it? Philosophers have tried to answer both questions by appealing to low mood’s phenomenal character. However, appeals to phenomenology have not settled this debate. Thus, I take a different approach: I tackle both questions by examining low mood’s complex functional role in cognition. I argue that if we take this role into account, we have excellent reason to believe that low mood a) has content, and b) has the following indicative-imperative content: Good events are, on average, less likely to occur than bad events & Limit [the subject’s] resource expenditure!

This paper advances a theory of disease as domino dysfunction. It is often argued that diseases are biological dysfunctions. However, a theory of disease as biological dysfunction is complicated by some plausible cases of dysfunction, which seem clearly non-pathological. I argue that pathological conditions are not just dysfunctions but domino dysfunctions, and that domino dysfunctions can be distinguished on principled biological grounds from non-pathological dysfunctions. I then show how this theory can make sense of the problem cases; they are not diseases because they are not domino dysfunctions.

Recent developments in foundations of physics have given rise to a class of views suggesting that physically meaningful descriptions must always be relativized to a physical perspective. In this article I distinguish between strong physical perspectivalism, which maintains that all facts must be relativized to a perspective, and moderate physical perspectivalism, which maintains that all empirically meaningful descriptions must be relativized to a perspective. I argue that scientific evidence and philosophical considerations support moderate physical perspectivalism over strong physical perspectivalism. In particular, motivations connected to epistemic humility and the social nature of science are more compatible with the moderate approach.

In response to the DSM’s crisis in validity, psychiatry has seen a proliferation of alternative research frameworks (HiTOP, the Network Approach, RDoC) for studying and classifying psychiatric disorders. In this paper, I argue the existence of multiple frameworks in which each employs their own standards of validity is problematic methodologically speaking for trying to do any kind of unified validation work. Fundamental disagreements concerning the underlying phenomenon, sources of validating evidence, and the very nature of validity move each framework into an unrecognized plurality. The consequence for psychiatry is a new validity crisis.