We use a theorem from M. J. Schervish to explore the relationship between accuracy and practical success. If an agent is pragmatically rational, she will quantify the expected loss of her credence with a strictly proper scoring rule. Which scoring rule is right for her will depend on the sorts of decisions she expects to face. We relate this pragmatic conception of inaccuracy to the purely epistemic one popular among epistemic utility theorists.

According to Michael Friedman’s theory of explanation, a law X explains laws Y1, Y2, …, Yn precisely when X unifies the Y’s, where unification is understood in terms of reducing the number of independently acceptable laws. Philip Kitcher criticized Friedman’s theory but did not analyze the concept of independent acceptability. Here we show that Kitcher’s objection can be met by modifying an element in Friedman’s account. In addition, we argue that there are serious objections to the use that Friedman makes of the concept of independent acceptability.

This article investigates a problem for statistical model evaluation, in particular for curve fitting: by employing a different family of curves we can fit any scatter plot almost perfectly at apparently minor cost in terms of model complexity. The problem is resolved by an appeal to prior probabilities. This leads to some general lessons about how to approach model evaluation.

Robustness arguments hold that hypotheses are more likely to be true when they are confirmed by diverse kinds of evidence. Robustness arguments require the confirming evidence to be independent. We identify two kinds of independence appealed to in robustness arguments: ontic independence (OI)—when the multiple lines of evidence depend on different materials, assumptions, or theories—and probabilistic independence. Many assume that OI is sufficient for a robustness argument to be warranted. However, we argue that, as typically construed, OI is not a sufficient independence condition for warranting robustness arguments. We show that OI evidence can collectively confirm a hypothesis to a lower degree than individual lines of evidence, contrary to the standard assumption undergirding usual robustness arguments. We employ Bayesian networks to represent the ideal empirical scenario for a robustness argument and a variety of ways in which empirical scenarios can fall short of this ideal.

This article investigates the relationship between two evaluative claims about agents’ degrees of belief: (i) that it is better to have more rather than less accurate degrees of belief and (ii) that it is better to have less rather than more probabilistically incoherent degrees of belief. We show that, for suitable combinations of inaccuracy measures and incoherence measures, both claims are compatible, although not equivalent; moreover, certain ways of becoming less incoherent always guarantee improvements in accuracy. Incompatibilities between particular incoherence and inaccuracy measures can be exploited to argue against particular ways of measuring either inaccuracy or incoherence.

Sober develops an account of adaptations on which they must have been positive causal factors for reproductive success. Glymour defends an account of a proper subset of adaptations—adaptations to particular environmental conditions—on which traits must interact in a special way with adapting conditions to cause reproductive success. These theories render conflicting judgments about which traits count as adaptations in some interesting cases. In this article I explore one such case and argue that we ought to replace the notion of adaptation qua positive causal factor with a novel notion of adaptation.

According to the Fine-Tuning Argument (FTA), the existence of life in our universe confirms the Multiverse Hypothesis (HM). A standard objection to FTA is that it violates the Requirement of Total Evidence (RTE). I argue that RTE should be rejected in favor of the Predesignation Requirement, according to which, in assessing the outcome of a probabilistic process, we should only use evidence characterizable in a manner available before observing the outcome. This produces the right verdicts in some simple cases in which RTE leads us astray, and, when applied to FTA, it shows that our evidence does confirm HM.

The first part of this article finds Craver’s mutual manipulability theory (MM) of constitution inadequate, as it definitionally ties constitution to the feasibility of ideal experiments, which, however, are unrealizable in principle. As an alternative, the second part develops an abductive theory of constitution (NDC), which exploits the fact that phenomena and their constituents are unbreakably coupled via common causes. The best explanation for this fact is the existence of an additional dependence relation, namely, constitution. NDC has important ramifications for constitutional discovery—most notably, that there is no experimentum crucis for constitution, not even under ideal discovery circumstances.

A certain type of counterfactual is thought to be intimately related to causation, control, and explanation. The time asymmetry of these phenomena therefore plausibly arises from a time asymmetry of counterfactual dependence. But why is counterfactual dependence time asymmetric? The most influential account of the time asymmetry of counterfactual dependence is David Albert’s account, which posits a new, time-asymmetric fundamental physical law, the so-called past hypothesis. Albert argues that the time asymmetry of counterfactual dependence arises from holding fixed the past hypothesis when evaluating counterfactuals. In this article, I argue that Albert’s account misconstrues the time asymmetry of counterfactual dependence.

Competition between scientific hypotheses is not always a matter of mutual exclusivity. Consistent hypotheses can compete to varying degrees either directly or indirectly via a body of evidence. We motivate and defend a particular account of hypothesis competition by showing how it captures these features. Computer simulations of Bayesian inference are used to highlight the limitations of adopting mutual exclusivity as a simplifying assumption to model scientific reasoning, particularly due to the exclusion of hypotheses that may be true. We end with a case study demonstrating the subtleties involved in hypothesis competition in scientific practice.

A high heritability estimate usually corresponds to a situation in which trait variation is largely caused by genetic variation. However, in some cases of gene-environment covariance, causal intuitions about the sources of trait difference can vary, leading experts to disagree as to how the heritability estimate should be interpreted. We argue that the source of contention for these cases is an inconsistency in the interpretation of the concepts ‘genotype’, ‘phenotype’, and ‘environment’. We propose an interpretation of these terms under which trait variance initially caused by genetic variance is subsumed into a heritability for all cases of gene-environment covariance.

This article argues that not only are there serious internal difficulties with both Garber’s and later ‘Garber-style’ solutions of the old-evidence problem, including a recent proposal of Hartmann and Fitelson, but Garber-style approaches in general cannot solve the problem. It also follows the earlier lead of Rosenkrantz in pointing out that, despite the appearance to the contrary which inspired Garber’s nonclassical development of the Bayesian theory, there is a straightforward, classically Bayesian, solution.

This article examines Newton’s method of induction and its connection to methodological atomism. The article argues that Newton’s Rule III for the Study of Natural Philosophy is a criterion for isolating the primary qualities of the atomic parts; in other words, it interprets Rule III as a transductive inference. It is shown that both the standard inductive and invariance interpretations of Rule III can be subsumed under the transductive view, although the invariance criterion is reinterpreted; by qualities “that cannot be intended or remitted,” Newton means qualities that are invariant under changes in the configuration of the atomic parts.

The no free lunch theorem is a radicalized version of Hume’s induction skepticism. It asserts that relative to a uniform probability distribution over all possible worlds, all computable prediction algorithms—whether ‘clever’ inductive or ‘stupid’ guessing methods (etc.)—have the same expected predictive success. This theorem seems to be in conflict with results about meta-induction. According to these results, certain meta-inductive prediction strategies may dominate other (non-meta-inductive) methods in their predictive success (in the long run). In this article this conflict is analyzed and dissolved, by means of probabilistic analysis and computer simulation.

We show that previous results from epistemic network models by Kevin J. S. Zollman and Erich Kummerfeld showing the benefits of decreased connectivity in epistemic networks are not robust across changes in parameter values. Our findings motivate discussion about whether and how such models can inform real world epistemic communities.

Most attempts to define culture as used in the cultural evolution literature treat culture as a single phenomenon that can be given a single nondisjunctive definition. In this article I argue that, really, cultural evolutionists employ a variety of distinct but closely related concepts of culture. I show how the main prominent attempts to define a culture concept fail to properly capture all the uses of “culture” employed in cultural evolutionary work. I offer a description of some of the most important culture concepts used by cultural evolutionists.

This article suggests a revision of the theory of causal nets (TCN). In section 1 we introduce an axiomatization of TCN based on a realistic understanding. It is shown that the causal Markov condition entails three independent principles. In section 2 we analyze indeterministic decay as the major counterexample to one of these principles: screening off by common causes (SCC). We call SCC-violating common causes interactive causes. In section 3 we develop a revised version of TCN, called TCN*, which accounts for interactive causes. It is shown that there are interactive causal models that admit of no faithful noninteractive reconstruction.