First, I show that the semantic thesis of scientific realism may be relaxed significantly—to allow that some scientific discourse is not truth-valued—without making any concessions concerning the epistemic or methodological theses that lie at realism’s core. Second, I illustrate how relaxing the semantic thesis allows realists to avoid positing abstract entities and to fend off objections to the “no miracles” argument from positions such as cognitive instrumentalism. Third, I argue that the semantic thesis of scientific realism should be relaxed because it is possible for scientific statements to be partially true, and hence approximately true, without being false.

This paper explores a distinction among causal relationships that has yet to receive attention in the philosophical literature, namely, whether causal relationships are reversible or irreversible. We provide an analysis of this distinction and show how it has important implications for causal inference and modeling. This work also clarifies how various familiar puzzles involving preemption and over-determination play out differently depending on whether the causation involved is reversible.

The task of Philosophy in Science (PinS) is to use philosophical tools to help solve scientific problems. This article describes how I stumbled into this line of work and then addressed several topics in philosophy of biology—units of selection, cladistic parsimony, robustness and trade-offs in model building, adaptationism, and evidence for common ancestry—often in collaboration with scientists. I conclude by offering advice for would-be PinS practitioners.

Synthetic biology has a strong modal dimension that is part and parcel of its engineering agenda. In turning hypothetical biological designs into actual synthetic constructs, synthetic biologists reach toward potential biology instead of concentrating on naturally evolved organisms. We analyze synthetic biology’s goal of making biology easier to engineer through the combinatorial theory of possibility, which reduces possibility to (re)combinations of individuals and their attributes in the actual world. While the last decades of synthetic biology explorations have shown biology to be much more difficult to engineer than originally conceived, synthetic biology has not given up its combinatorial approach.

Models typically have actual, existing targets. However, some models are viewed as having non-actual targets. I argue that this interpretation comes at various costs and propose an alternative that fares better along two dimensions: (1) agreement with practice and (2) ontological and epistemological parsimony. My proposal is that many of these models actually have actual targets.

In analogical reasoning, observations about one or more source domains provide varying degrees of support for a conjecture about a target domain. Norton (2021) challenges the usefulness of formal models of analogical inference. Other philosophers (Dardashti et al. 2019) develop just such formal models in order to show how analogue experiments can confirm a hypothesis, even when the target domain is inaccessible. This paper defends the value of quasi-formal models of analogical reasoning. Such models are broadly compatible with Norton’s position, but help to clarify the structure of analogical reasoning and to identify basic requirements for a good analogical inference.

Laurie Paul argues that, when it comes to many of your most significant life-changing decisions, the principles of rational choice are silent. That is because, in these cases, you anticipate that one of your choice options would yield a transformative experience. We argue that such decisions are best seen as ones in which you anticipate awareness growth. You do not merely lack knowledge about which possible outcome will arise from a transformative option; you lack knowledge about what are the possible outcomes. We show how principles of rational choice can be extended to cases of anticipated awareness growth.

It is a plausible speculation that conventional choices in outcome measures might influence the results of meta-analyses. We test that speculation by simulating data from trials on antidepressants. We vary real drug effectiveness while modulating conventional values for outcome measures. We had previously shown that one conventional choice used in meta-analyses of antidepressants falls in a narrow range of values that maximize estimates of effectiveness. Our present analysis investigates why this phenomenon occurs. Moreover, our results suggest the superiority of absolute outcome measures over relative measures. This research program can be extended to test numerous other aspects of clinical research.

The role or function of experimental and observational replication within empirical science has implications for how replication should be measured. Broadly, there seems to be consensus that replication’s central goal is to confirm or vouchsafe the reliability of scientific findings. I argue that if this consensus is correct, then most of the measures of replication used in the scientific literature are actually poor indicators of this reliability or confirmation. Only meta-analytic measures of replication align functionally with the goals of replication. I conclude by addressing some objections to meta-analysis.

Numerous philosophers of science have argued that the incentive structure of science is a major contributor to the replication crisis. In this article I review FiveThirtyEight’s election forecasting model and show that it confronts similar problems in political polling and successfully mitigates them, not by changing the incentives, but by weighting a pollster’s effect on the model by their past reliability. I argue that a similar weighting procedure holds a promising approach to addressing the replication crisis in science.

Philosophical work on meta-analysis focuses on biomedical research and revolves around the question: Is meta-analytic evidence the best kind of evidence? By focusing on conservation science rather than biomedical science, I identify further questions and puzzles for meta-analysis and show their importance for the epistemology of meta-analysis.

I argue that Open Science as currently conceptualized and implemented does not take sufficient account of epistemic diversity within research. I use three case studies to exemplify how Open Science threatens to privilege some forms of inquiry over others, thus exasperating divides within and across systems of practice, and overlooking important sources and forms of epistemic diversity. Building on insights from pluralist philosophy, I then identify four aspects of diverse research practices that should serve as reference points for debates around Open Science: (1) specificity to local conditions, (2) entrenchment within repertoires, (3) permeability to newcomers, and (4) demarcation strategies.

Merton envisioned his norms of science at a time when peer-reviewed journals controlled scientific communication. Technologies for sharing and finding content have since divorced the certification and amplification of science, generating systemic vulnerabilities. Certified amplification—a new Mertonian-styled norm—enjoins their recoupling and introduces a taxonomy of strategies adopted by institutions to close the certification-amplification gap, including the proportioning of the one to the other. Examples illustrating each taxonomic type collectively paint a picture of an ethos employing a rich range of certification and amplification techniques and emerging in a decentralized fashion across heterogeneous objects, communication modalities, and institutions.

A major goal of the open science movement is to make more scientific information available to non-specialists, but it has been difficult to meaningfully achieve that goal. In response, this paper argues for two steps: (1) focusing on the scientific content that is most relevant to non-specialist audiences; and (2) packaging that content in meaningful ways for those audiences. The paper uses a case study involving a major environmental health issue (namely, PFAS pollution) to illustrate how the proponents of open science can work with groups like government agencies, nongovernmental organizations, and extension services to implement these two steps.

The prospective formalization of the Anthropocene as a chronostratigraphic unit by the International Commission on Stratigraphy has been intensely debated. This paper explores and assesses the stakes of this process from a philosophical perspective. I distinguish two senses of formalization—the descriptive and the evaluative—and argue that: 1) there are descriptive and evaluative formalizations of the Anthropocene beyond the confines of the ICS; 2) incoherencies between Anthropocene proposals and the ICS’s current tenets are not a decisive reason for deferring descriptive formalization; and 3) the prospective evaluative formalization of the Anthropocene by the ICS would be impactful.

Scientists are frequently called upon to “democratize” science, by bringing the public into scientific research. One appealing point for public involvement concerns the nonepistemic values involved in science. Suppose, though, a scientist invites the public to participate in making such value-laden determinations but finds that the public holds values the scientist considers morally unacceptable. Does the argument for democratizing science commit the scientist to accepting the public’s objectionable values, or may she veto them? I argue that there are a limited set of cases in which scientists can, consistently with a commitment to democratized science, set aside the public’s judgments.

The COVID-19 pandemic and global climate change crisis remind us that widespread trust in the products of the scientific enterprise is vital to the health and safety of the global community. Insofar as appropriate responses to these (and other) crises require us to trust that enterprise, cultivating a healthier trust relationship between science and the public may be considered as a collective public good. While it might appear that scientists can contribute to this good by taking more initiative to communicate their work to public audiences, we raise a concern about unintended consequences of an individualistic approach to such communication.

Transparency has been proposed as a way of handling value-ladenness in machine learning (ML). This article highlights limits to this strategy. I distinguish three kinds of transparency: epistemic transparency, retrospective value transparency, and prospective value transparency. This corresponds to different approaches to transparency in ML, including so-called explainable artificial intelligence and governance based on disclosing information about the design process. I discuss three sources of value-ladenness in ML—problem formulation, inductive risk, and specification gaming—and argue that retrospective value transparency is only well-suited for dealing with the first, while the third raises serious challenges even for prospective value transparency.

Under what conditions does machine learning (ML) model opacity inhibit the possibility of explaining and understanding phenomena? In this article, I argue that nonepistemic values give shape to the ML opacity problem even if we keep researcher interests fixed. Treating ML models as an instance of doing model-based science to explain and understand phenomena reveals that there is (i) an external opacity problem, where the presence of inductive risk imposes higher standards on externally validating models, and (ii) an internal opacity problem, where greater inductive risk demands a higher level of transparency regarding the inferences the model makes.