Philosophically, one of the most important questions in the enterprise termed confirmation theory is this: Why should one stick to well confirmed theories rather than to any other theories? This paper discusses the answers to this question one gets from absolute and incremental Bayesian confirmation theory. According to absolute confirmation, one should accept “absolutely well confirmed” theories, because absolute confirmation takes one to true theories. An examination of two popular measures of incremental confirmation suggests the view that one should stick to incrementally well confirmed theories, because incremental confirmation takes one to (the most) informative (among all) true theories. However, incremental confirmation does not further this goal in general. I close by presenting a necessary and sufficient condition for revealing the confirmational structure in almost every world when presented separating data.

I present a game-theoretic way to understand the situation describing Newcomb's Problem (NP) which helps to explain the intuition of both one-boxers and two-boxers. David Lewis has shown that the NP may be modelled as a Prisoner's Dilemma game (PD) in which ‘cooperating’ corresponds to ‘taking one box’. Adopting relevant results from game theory, this means that one should take just one box if the NP is repeated an indefinite number of times, but both boxes if it is a one-shot game. Causal decision theorists thus give the right answer for the one-shot situation, whereas the one-boxers’ solution applies to the indefinitely iterated case. Because Nozick's set-up of the NP is ambiguous between a one-shot and a repeated game, both of these solutions may appear plausible—depending on whether one conceives of the situation as one-off or repeated. If the players’ aim is to maximize their payoffs, the symmetric structure of the PD implies that the two players will behave alike both when the game is one-shot and when it is played repeatedly. Therefore neither the observed outcome of both players selecting the same strategy (in the PD) nor, correspondingly, the predictor's accurate prediction of this outcome (in the NP) is at all surprising. There is no need for a supernatural predictor to explain the NP phenomena.

This paper addresses the problem that Bayesian statistical inference cannot accommodate theory change, and proposes a framework for dealing with such changes. It first presents a scheme for generating predictions from observations by means of hypotheses. An example shows how the hypotheses represent the theoretical structure underlying the scheme. This is followed by an example of a change of hypotheses. The paper then presents a general framework for hypotheses change, and proposes the minimization of the distance between hypotheses as a rationality criterion. Finally the paper discusses the import of this for Bayesian statistical inference.

A variety of robustness objections have been made against evolutionary game theory. One of these objections alleges that the games used in the underlying model are too arbitrary and oversimplified to generate a robust model of interesting prosocial behaviors. In this paper, I argue that the robustness objection can be met. However, in order to do so, we must attend to important conceptual issues regarding the nature of fairness, justice, and other moral concepts. Specifically, we must better understand the relationship between moral concepts and formal characterizations of games.

Studies in the social and biomedical sciences of racial differences in socioeconomic status or health within a population view the race of members as fixed and look for a difference in the frequency of a trait like average income or disease risk between racial subgroups. But, as I explain in this paper, there are good reasons to allow the race of members to vary with the trait whose variation within the population is to be described or explained. According to such a view of race, racial categories are more scientifically significant if membership in the categories is allowed to vary with differences in scientific interest rather than held constant across a variety of interests.

There has been little serious work to integrate the constructionist approach and the cognitive/evolutionary approach in the domain of race, although many researchers have paid lip service to this project. We believe that any satisfactory account of human beings’ racialist cognition has to integrate both approaches. In this paper, we propose to move toward this integration. We present an evolutionary hypothesis that rests on a distinction between three kinds of groups—kin-based groups, small scale coalitions, and ethnies. Following Gil-White (1999, 2001a, 2001b), we propose that ethnies have raised specific evolutionary challenges that were solved by an evolved cognitive system. We suggest that the concept of race is a byproduct of this mechanism. We argue that recent theories of cultural transmission are our best hope for integrating social constructionists’ and cognitive/evolutionary theorists’ insights.

Some thinkers distrust Darwinist explorations of complex human behaviors, particularly investigations into possible differences in valued skills between genders, races or classes. Such projects, it is claimed, tend to have adverse effects on people who are already disadvantaged. A recent argument by Philip Kitcher both clarifies and generalizes this charge to cover a wider genre of scientific projects. In this paper I try to spell out and analyze Kitcher's argument. The argument fails, I suggest, because some of its key premises fail to convince. My analysis focuses on relevant facts about the role of inquiry in fallibilist contexts, the texture of belief in contemporary natural science, and the moral dimension of scientific research.

Current controversy over whether the Office of Management and Budget (OMB) system of racial and ethnic classification should be used in pharmacogenetics research as suggested by the U.S. Federal Drug Administration (FDA) has been couched in terms of realist-social constructionist debates on race. The assumptions both parties to these debates share instead need to be relinquished—specifically, dichotomies between the social and scientific and what is descriptive and evaluative/normative. This paper defends a pragmatic approach to the question of the appropriateness of the OMB group categories in pharmacogenetics research, an approach that is local and context-specific rather than global, incorporates practical and ethical as well as theoretical dimensions, and recognizes intersections of the social and the biological in the constitution of group categories.

The partial structures program of da Costa, French and others offers a unified framework within which to handle a wide range of issues central to contemporary philosophy of science. I argue that the program is inadequately equipped to account for simple cases where idealizations are used to construct abstract, mathematical models of physical systems. These problems show that da Costa and French have not overcome the objections raised by Cartwright and Suárez to using model-theoretic techniques in the philosophy of science. However, my concerns arise independently of the more controversial assumptions that Cartwright and Suárez have employed.

Representation has been one of the main themes in the recent discussion of models. Several authors have argued for a pragmatic approach to representation that takes users and their interpretations into account. It appears to me, however, that this emphasis on representation places excessive limitations on our view of models and their epistemic value. Models should rather be thought of as epistemic artifacts through which we gain knowledge in diverse ways. Approaching models this way stresses their materiality and media-specificity. Focusing on models as multifunctional artifacts releases them from any preestablished and fixed representational relationships and leads me to argue for a twofold approach to representation.

Loop analysis is a method of qualitative modeling anticipated by Sewall Wright (1921) and systematically developed by Richard Levins. In Levins’ (1966) distinctions between modeling strategies, loop analysis sacrifices precision for generality and realism. Besides criticizing the clarity of these distinctions, Orzack and Sober (1993) argued qualitative modeling is conceptually and methodologically problematic. Loop analysis of the stability of ecological communities shows this criticism is unjustified. It presupposes an overly narrow view of qualitative modeling and underestimates the broad role models play in scientific research, especially in helping scientists represent and understand complex systems.

I argue against the conception of scientific models advocated by the proponents of the Semantic View of scientific theories. Part of the paper is devoted to clarifying the important features of the scientific modeling view that the Semantic conception entails. The liquid drop model of nuclear structure is analyzed in conjunction with the particular auxiliary hypothesis that is the guiding force behind its construction and it is argued that it does not meet the necessary features to render it a model of the theory, as the Semantic View demands. Given that this model is indicative of how quantum mechanics is applied in the domain of nuclear physics, I claim that the Semantic View does not adequately account for scientific models.

I offer a novel argument for spacetime substantivalism: We should take the spacetime of general relativity to be a substance because of its active role in gravitational causation. As a clear example of this causal behavior I offer the cosmological constant, a term in the most general form of the Einstein field equations which causes free floating objects to accelerate apart. This acceleration cannot, I claim, be causally explained except by reference to spacetime itself.

Using Einstein's comparison of General Relativity and Descartes’ physics, this investigation explores the alleged conventionalism that pervades the ontology of substantival and relationist conceptions of spacetime. Although previously discussed, namely by Rynasiewicz and Hoefer, it will be argued that the close similarities between General Relativity and Cartesian physics have not been adequately treated in the literature—and that the disclosure of these similarities bolsters the case for a conventionalist interpretation of spacetime ontology.

Four techniques for measuring the masses of galaxies and larger astrophysical systems from their dynamics are discussed. Their apparent agreement is sometimes invoked as warrant for postulating huge quantities of ‘dark matter’ as the best solution to “the dynamical discrepancy,” the disparity between the amount of mass visible in large scale astrophysical systems and the amount calculated from dynamics. This paper argues that the agreement, though suggestive, is not definitive. The coincident measurements remain the best reason for preferring dark matter over revisions to General Relativity for solving the dynamical discrepancy, but the preference is only weakly warranted.

A proper understanding of black hole complementarity as a response to the information loss paradox requires recognizing the essential role played by arguments for the applicability and limitations of effective semiclassical theories. I argue that this perspective sheds important light on the arguments advanced by Susskind, Thorlacius, and Uglum—although ultimately I argue that their position is unsatisfactory. I also consider the argument offered by 't Hooft for the breakdown of microcausality around black holes, and conclude that it relies on a mistaken treatment of measurement collapse. There is, however, a legitimate argumentative role for black hole complementarity, exemplified by the position of Kiem, Verlinde, and Verlinde, that calls for a more subtle analysis of the limitations facing our effective theories.

In an attempt to avert Laudan's pessimistic induction, Kitcher, Worrall and Psillos introduce a narrower version of scientific realism. According to this version, which can be referred to as “localized realism,” realists need not accept every component in a successful theory. They are supposed only to accept those components that led to the theory's empirical success. Consequently, realists can avoid believing in dubious entities like the caloric and ether. This paper examines and critiques localized realism. It also scrutinizes Psillos's historical study of the caloric theory of heat, which is intended to support localized realism.

In a recent PSA paper (2001a), as well as some other papers (1995, 2000, 2001b) and a book chapter (1999, Chapter 7), Stathis Psillos raised a number of objections against structural realism. The aim of this paper is threefold: (1) to evaluate part of Psillos’ offence on the Russellian version of epistemic structural realism (ESR); (2) to elaborate more fully what Russellian ESR involves; and (3) to suggest improvements where it is indeed failing.

John Worrall recently provided an account of epistemic structural realism, which explains the success of science by arguing for the correct mathematical structure of our theories. He accounts for the historical failures of science by pointing to bloated ontological interpretations of theoretical terms. In this paper I argue that Worrall's account suffers from five serious problems. I also show that Pierre Cruse and David Papineau have developed a rival structural realism that solves all of the problems faced by Worrall. This Ramsey sentence realism is a significant advance in the debate, but still ultimately fails for its incomplete account of reference

I argue that the pessimistic meta-induction (PMI) seems to point an ontological priority of the relations over the objects of the scientific theories of the kind suggested by French and Ladyman (2003). My strategy will involve a critical examination of epistemic structural realism (ESR) and an historical case-study: the prediction of Zeeman's effect in Lorentz's theory of the electron.