Some scientific realists suggest that scientific communities have improved in their ability to discover alternative theoretical possibilities and that the problem of unconceived alternatives therefore poses a less significant threat to contemporary scientific communities than it did to their historical predecessors. I first argue that the most profound and fundamental historical transformations of the scientific enterprise have actually increased rather than decreased our vulnerability to the problem. I then argue that whether we are troubled by even the prospect of increasing theoretical conservatism in science should depend on the position we occupy in the ongoing debate concerning scientific realism itself.

The Doomsday argument and anthropic reasoning are two puzzling examples of probabilistic confirmation. In both cases, a lack of knowledge apparently yields surprising conclusions. Since they are formulated within a Bayesian framework, they constitute a challenge to Bayesianism. Several attempts, some successful, have been made in a Bayesian framework that represents credal states by single credence functions to avoid these conclusions, but none of them can do so for all versions of the Doomsday argument. I show that adopting an imprecise framework of probabilistic reasoning allows for a more adequate representation of ignorance and explains away these puzzles.

Stanley Stevens draws a useful distinction among ordinal scales, interval scales, and ratio scales. Most recent discussions of confirmation measures have proceeded on the ordinal level of analysis. In this article, I give a more quantitative analysis. In particular, I show that the requirement that our desired confirmation measure be at least an interval measure naturally yields necessary conditions that jointly entail the log-likelihood measure. Thus, I conclude that the log-likelihood measure is the only good candidate interval (or ratio) measure.

Recent claims, mainly from computer scientists, concerning a largely automated and model-free data-intensive science have been criticized by several philosophers of science. The debate suffers from lack of detail regarding the actual methods used in data-intensive science and in which ways these presuppose theoretical assumptions. I examine two widely used algorithms, classificatory trees and nonparametric regression, and argue that they are theory laden in an external sense, regarding the framing of research questions, but not in an internal sense, concerning the causal structure of the examined phenomenon. With respect to the novelty of data-intensive science, I draw an analogy to exploratory experimentation.

“Dutch Book” arguments and references to gambling theorems are typical in the debate between Bayesians and scientists committed to “classical” statistical methods. These arguments have rarely convinced non-Bayesian scientists to abandon certain conventional practices (like fixed-level null hypothesis significance testing), partially because many scientists feel that gambling theorems have little relevance to their research activities. In other words, scientists “don’t bet.” This article examines one attempt, by Schervish, Seidenfeld, and Kadane, to progress beyond such apparent stalemates by connecting “Dutch Book”–type mathematical results with principles actually endorsed by practicing experimentalists.

Forster presented some interesting examples having to do with distinguishing the direction of causal influence between two variables, which he argued are counterexamples to the likelihood theory of evidence. In this article, we refute Forster’s arguments by carefully examining one of the alleged counterexamples. We argue that the example is not convincing as it relies on dubious intuitions that likelihoodists have forcefully criticized. More important, we show that contrary to Forster’s contention, the consilience-based methodology he favored is accountable within the framework of the likelihood theory of evidence.

Modern mathematical sciences are hard to imagine without appeal to efficient computational algorithms. We address several conceptual problems arising from this interaction by outlining rival but complementary perspectives on mathematical tractability. More specifically, we articulate three alternative characterizations of the complexity hierarchy of mathematical problems that are themselves based on different understandings of computational constraints. These distinctions resolve the tension between epistemic contexts in which exact solutions can be found and the ones in which they cannot; however, contrary to a persistent myth, we conclude that having an exact solution is not generally more epistemologically beneficial than lacking one.

In recent years the social nature of scientific inquiry has generated considerable interest. We examine the effect of an epistemically impure agent on a community of honest truth seekers. Extending a formal model of network epistemology pioneered by Zollman, we conclude that an intransigently biased agent prevents the community from ever converging to the truth. We explore two solutions to this problem, including a novel procedure for endogenous network formation in which agents choose whom to trust. We contend that our model nicely captures aspects of current problems in medical research and gesture at some morals for medical epistemology more generally.

Cheap talk has often been thought incapable of supporting the emergence of cooperation because costless signals, easily faked, are unlikely to be reliable. I show how, in a social network model of cheap talk with reinforcement learning, cheap talk does enable the emergence of cooperation, provided that individuals also temporally discount the past. This establishes one mechanism that suffices for moving a population of initially uncooperative individuals to a state of mutually beneficial cooperation even in the absence of formal institutions.

I investigate the structure of an argument that culture cannot be maintained in a population if each individual learns only from a single person. This appears to conflict with (1) many models of cultural transmission and (2) real-world cases. I resolve the first problem by showing that one of the models central to the argument is conceptually analogous and mathematically equivalent to one used to investigate the evolution of sexual reproduction. I resolve the second by arguing that probabilistic models of epistemological coherence can be reinterpreted as models of support between cultural variants—illustrating the idea with a new model.

The handicap principle is one of the most influential ideas in evolutionary biology. It asserts that when there is conflict of interest in a signaling interaction signals must be costly in order to be reliable. While in evolutionary biology it is a common practice to distinguish between indexes and fakable signals, we argue this dichotomy is an artifact of existing popular signaling models. Once this distinction is abandoned, we show one cannot adequately understand signaling behavior by focusing solely on cost. Under our reframing, cost becomes one—and probably not the most important—of a collection of factors preventing deception.

Multifunctionality poses significant challenges for human brain mapping. Cathy Price and Karl Friston argue that brain regions perform many functions in one sense and a single function in another. Thus, neuroscientists must revise their “cognitive ontologies” to obtain systematic mappings. Colin Klein draws a different lesson from these findings: neuroscientists should abandon systematic mappings for context-sensitive ones. I claim that neither account succeeds as a general treatment of multifunctionality. I argue that brain areas, like genes or organs, are multifunctional in different ways. I call this the “functional heterogeneity hypothesis.” I contend that different multifunctional parts require different mapping strategies.

Current theories of addiction try to explain what addiction is, who experiences it, why it occurs, and how it develops and persists. In this article, I explain why none of these theories can be accepted as a comprehensive model. I argue that current models fail to account for differences in embodiment, interaction processes, and the experience of addiction. To redress these limiting factors, I design a proposal for an enactive account of addiction that follows the enactive model of autism proposed by Hanne De Jaegher.

Mechanistic explanations satisfy widely held norms of explanation: the ability to manipulate and answer counterfactual questions about the explanandum phenomenon. A currently debated issue is whether any nonmechanistic explanations can satisfy these explanatory norms. Weiskopf argues that the models of object recognition and categorization, JIM, SUSTAIN, and ALCOVE, are not mechanistic yet satisfy these norms of explanation. In this article I argue that these models are mechanism sketches. My argument applies recent research using model-based functional magnetic resonance imaging, a novel neuroimaging method whose significance for current debates on psychological models and mechanistic explanation has yet to be explored.

Psychiatry’s failure to validate its diagnostic constructs is often attributed to the prioritizing of reliability over validity in the Diagnostic and Statistical Manual of Mental Disorders (DSM). I argue that a more powerful way in which the DSM has retarded biomedical progress is by encouraging unwarranted optimism about diagnostic discrimination: the assumption that our diagnostic tests group patients together in ways that allow for relevant facts about mental disorder to be discovered. I argue that the Research Domain Criteria (RDoC) framework, a new paradigm for classifying objects of psychiatric research, solves some of the challenges brought on by this assumption.

In guarding against inferential mistakes, experimental comparative cognition errs on the side of underattributing sophisticated cognition to animals, or what I refer to as the underattribution bias. I propose eliminating this bias by altering the method of choosing the default, or null, hypothesis. Rather than choosing the most parsimonious null hypothesis, as is current practice, I argue for choosing the best-evidenced hypothesis. Doing so at once preserves the risk-controlling structure of the current statistical paradigm and introduces a sensitivity to probability-conferring empirical and theoretical information. This analysis illustrates how values like parsimony can covertly shape statistical-experimental design and inference.