Mixture of experts is a prediction aggregation method in machine learning that aggregates the predictions of specialized experts. This method often outperforms Bayesian methods despite the Bayesian having stronger inductive guarantees. We argue that this is due to the greater functional capacity of mixture of experts. We prove that in a limiting case, mixture of experts will have greater capacity than equivalent Bayesian methods, which we vouchsafe through experiments on non-limiting cases. Finally, we conclude that mixture of experts is a type of abductive reasoning in the Peircean sense of hypothesis construction.

Statistical criteria of fairness, although controversial, bring attention to the multiobjective nature of many predictive modeling problems. In this article, I consider how epistemic and nonepistemic values help to justify the design of machine learning algorithms that optimize for more than one normative goal. I focus on a major design choice between biased search strategies that directly incorporate priorities for various objectives into an optimization procedure and unbiased search strategies that do not. I argue that both reliably generate Pareto optimal solutions such that various other values are relevant to making a rational choice between them.

Galaxy clusters are commonly used tracers of cosmology. Gravitational lensing analysis of the Bullet Cluster is claimed to evidentially support dark matter, an important component in the ${\rm{\Lambda }}$CDM cosmology. I argue that such ${\rm{\Lambda }}$CDM-based models of individual galaxy clusters should be explanatory to meet such claims, but hardly in an ontic sense, due to galaxy cluster anisotropies, empirically equivalent non-${\rm{\Lambda }}$CDM-based models, and currently unaccountable cases. I propose that adopting an alternative epistemic/representational conception of scientific explanation can maintain the explanatory nature of individual galaxy cluster models, cope with the three complications, and be potentially generalizable to other branches of astrophysics.

The stable property cluster view of natural kinds leaves the metaphysical structure of natural kinds open. The positive corollary of this feature is freedom from explaining the nature of problematic entities and that it affords natural kind status and flexibility. However, some worries are raised against this feature; the metaphysical structure of natural kinds is required for explaining their epistemic values such as inductive and explanatory power. Here I take the criticisms at face value and offer a possible metaphysical model of the SPC account that does not meddle with the merits but mitigates the worries.

This paper discusses the role of data within scientific reasoning and as evidence for theoretical claims, arguing for the idea that data can yield theoretically grounded models and be inferred, predicted, or explained from/by such models. Contrary to Bogen and Woodward’s rejection of data-to-theory and theory-to-data inferences/predictions, we draw upon artificial intelligence as applied to science literature to argue that (a) many models are routinely inferred and predicted from the data and routinely used to infer and predict data, and (b) such models can, at least in some contexts, play the role of theories.

This paper challenges the notion of emergent time in quantum cosmology by examining the reconciliation of the timeless Wheeler–DeWitt equation with the Universe’s dynamical evolution. We critically evaluate the analogy between the Wheeler–DeWitt and Klein–Gordon equations, highlighting challenges for the identification of an emergent time parameter. We conclude that refining this analogy may lead to a better understanding of emergent time in quantum cosmology, though it is still not free from complications.

Recent work by Michael O. Hardimon and Quayshawn Spencer defends a minimalist (or deflationary) biological realism about race. Their approach has two distinct features. First, unlike revisionist biological race, minimalist biological races are a conception of race that correspond to our ordinary race concepts. Second, unlike hereditarian or essentialist accounts, minimalist biological races are not claimed to be robustly explanatory. This article argues against their account of the biological genuineness of race. I argue that the minimalist biological conception of race lacks the explanatory constraints of genuine biological kinds. Rather, minimalist biological races are gerrymandered kinds.

Spacetime singularities are expected to disappear in quantum gravity. Singularity resolution prima facie supports the view that spacetime singularities are mathematical pathologies of general relativity. However, this conclusion might be premature. Spacetime singularities are more accurately understood as global properties of spacetime rather than things. Therefore, if spacetime emerges in quantum gravity—as it is often claimed—then so may its singular structure. Although this proposal is intriguing, the attempt to uphold that spacetime singularities may be emergent fails. I provide three arguments in support of this claim, drawing on different views of spacetime emergence.

The timing of the origin of flowering plants (Angiosperm) is hotly debated. It has been suggested that the disagreement between the fossil record of angiosperm origin strongly conflicts with the origin estimates generated by molecular clocks. I argue that this conflict reveals lessons about whether or under what conditions scientific disagreement is likely to bear fruit. Specifically, I point to issues of evidence quality and social epistemic structures which deserve more attention in understanding the productivity of disagreement.

It is often said that successful scientific research must be built on trust. Focusing on the alleged necessity of trust for successful scientific communication and thus for scientific cooperation (which underlies much of contemporary science), I argue that science mustn’t be built on trust. Appearances to the contrary come from a failure to distinguish different attitudes toward scientists’ testimony, in particular, trusting and relying on other scientists. This article proposes an account of scientific reliance and explains how it differs from scientific trust; it also shows why this distinction matters for science.

One challenge to relationism in general relativity is that the metric field is underdetermined by the stress–energy tensor. This is manifested in the existence of distinct vacuum solutions to Einstein’s equations. In this paper I reformulate the problem of underdetermination as a problem from vacuum solutions. I call this the vacuum challenge and identify the gravitational degrees of freedom (associated with the Weyl tensor) as the “source” of the challenge. The Weyl tensor allows for gravitational effects that something outside a system exerts on the system. I provide a relationist response to the vacuum challenge.

Thick concepts have both descriptive and evaluative dimensions to their meaning. Some have argued that because the descriptive and evaluative dimensions cannot be separated (they are “blended”), the implicit values influence the confirmation of any “mixed claims’’ containing the thick concept. Using the development of the concept of hypersegregation as a case study, we argue for a distinction between the semantic function of definitions and the epistemic function of indicators. While thick concepts are semantically blended, the evaluative meaning of a thick concept does not influence the function of epistemic indicators. Therefore, mixed claims can and should be tested impartially.

Advances in animal sentience research, neural organoids, and artificial intelligence reinforce the relevance of justifying attributions of consciousness to nonstandard systems. Clarifying the argumentative structure behind these attributions is important for evaluating their validity. This article addresses this issue, concluding that analogical abduction—a form of reasoning combining analogical and abductive elements—is the strongest method for extrapolating consciousness from humans to nonstandard systems. We argue that the argument from analogy and inference to the best explanation, individually taken, do not meet the criteria for successful extrapolations, whereas analogical abduction offers a promising approach despite limitations in current consciousness science.

Previous work on the evolution of the indicative–imperative distinction in signaling games has focused on either the cognitive processes used by signaling agents or information carried by messages. In this paper, I use a simple signaling game model to demonstrate that an overlooked feature of the indicative–imperative distinction can be learned. This overlooked feature is the fact that the indicative–imperative distinction is often marked in natural languages.

There is a growing consensus among philosophers that quantifying value-laden concepts can be epistemically successful and politically legitimate if all value-laden choices in the process of quantification are aligned with stakeholder values. I argue that proponents of this alignment approach have failed to argue for its basic premise: Successful quantification is sufficiently unconstrained to be achievable along multiple, stakeholder-specific pathways. I then challenge this premise by considering a rare example of successful value-laden quantification in seismology, in which stakeholder values had to be disregarded from measure design and testing. The example motivates my contention that value alignment is not a workable source of political legitimacy for successful programs of quantification.

In this article, I consider two arguments concerning the status of holobionts as evolutionary individuals—one rejects their status by privileging the “stability of lineages” and the other supports their status by privileging the “stability of traits.” I argue that the tension between these two arguments arises from two fundamentally different accounts of natural selection. I suggest that each account of selection corresponds to a unique account of evolutionary individuality. This strategy entitles us to a modest pluralism: Holobionts are evolutionary individuals on one account of selection but not on the other.

This article challenges a false dichotomy between subjectivity and objectivity in understanding the nature of human social relationships. I argue that social relationships are composed of both subjective and objective components, which are inherently interdependent. They are influenced by biological properties and subject to evolutionary processes, yet they cannot be reduced to them. I use emerging research on kinship as an example that showcases the appeal of this integrated approach. This article takes a step in the direction of a unified account of sociality, contributing to a more comprehensive understanding of human social behavior.

I use Swampman to illuminate the role of thought experiments in philosophy of science. Against Millikan and others, I argue that even outlandish thought experiments can shed light on science and scientific kinds, so long as we understand them as illustrations of scientific reasoning, not as examples of scientific kinds. The logic of thought experiments, understood as illustrations, is analogous to the logic of common experimental paradigms in science. So, in reviving Swampman and showing how he survives teleosemantic objections, I also provide a framework for understanding how, why, and when thought experiments are informative about science and scientific kinds.

In principle, inaccuracies in the representation of the climate’s internal variability could undermine the measurement of the human contribution to warming. Equally in principle, the success of the measurement practice could provide evidence that our assumptions about internal variability are correct. I argue that neither condition obtains: Current measurement practices do not provide evidence for the accuracy of our assumptions precisely because they are not as sensitive to inaccuracy in the representation of internal variability as might be worried. I end by drawing some lessons about stability and “robustness reasoning” more generally.

Subject to techniques of perturbative renormalization, the Standard Model makes empirical predictions that are stupendously successful. But also deeply mysterious. Not every quantum field theory (qft) is renormalizable. Indeed, most aren’t. The mystery is: Why should we be so lucky, that we live in a world governed by a renormalizable qft? I explicate this Renormalizability Puzzle, and explain why Renormalization Group (RG) approaches are widely thought to resolve it. Looking under the hood of the RG resolution, I identify a load-bearing element that might not be adequate to the explanatory burden the RG resolution places upon it.