A reprint of the Prentice-Hall edition of 1992. Prepared by nine distinguished philosophers and historians of science, this thoughtful reader represents a cooperative effort to provide an introduction to the philosophy of science focused on cultivating an understanding of both the workings of science and its historical and social context. Selections range from discussions of topics in general methodology to a sampling of foundational problems in various physical, biological, behavioral, and social sciences. Each chapter contains a list of suggested readings (...) and study questions. (shrink)

Recently, Rueger and Sharp (1996) and Koperski (1998) have been concerned to show that certain procedural accounts of model confirmation are compromised by non-linear dynamics. We suggest that the issues raised are better approached by considering whether chaotic data analysis methods allow for reliable inference from data. We provide a framework and an example of this approach.

This paper describes the application of eight statistical and machine-learning methods to derive computer models for predicting mortality of hospital patients with pneumonia from their findings at initial presentation. The eight models were each constructed based on 9847 patient cases and they were each evaluated on 4352 additional cases. The primary evaluation metric was the error in predicted survival as a function of the fraction of patients predicted to survive. This metric is useful in assessing a model’s potential to assist (...) a clinician in deciding whether to treat a given patient in the hospital or at home. We examined the error rates of the models when predicting that a given fraction of patients will survive. We examined survival fractions between 0.1 and 0.6. Over this range, each model’s predictive error rate was within 1% of the error rate of every other model. When predicting that approximately 30°K of the patients will survive, all the models have an error rate of less than 1.5%. The models are distinguished more by the number of variables and parameters that they contain than by their error rates; these differences suggest which models may be the most amenable to future implementation as paper-based guidelines. (shrink)

Measures of algorithmic bias can be roughly classified into four categories, distinguished by the conditional probabilistic dependencies to which they are sensitive. First, measures of "procedural bias" diagnose bias when the score returned by an algorithm is probabilistically dependent on a sensitive class variable (e.g. race or sex). Second, measures of "outcome bias" capture probabilistic dependence between class variables and the outcome for each subject (e.g. parole granted or loan denied). Third, measures of "behavior-relative error bias" capture probabilistic dependence between (...) class variables and the algorithmic score, conditional on target behaviors (e.g. recidivism or loan default). Fourth, measures of "score-relative error bias" capture probabilistic dependence between class variables and behavior, conditional on score. Several recent discussions have demonstrated a tradeoff between these different measures of algorithmic bias, and at least one recent paper has suggested conditions under which tradeoffs may be minimized. -/- In this paper we use the machinery of causal graphical models to show that, under standard assumptions, the underlying causal relations among variables forces some tradeoffs. We delineate a number of normative considerations that are encoded in different measures of bias, with reference to the philosophical literature on the wrongfulness of disparate treatment and disparate impact. While both kinds of error bias are nominally motivated by concern to avoid disparate impact, we argue that consideration of causal structures shows that these measures are better understood as complicated and unreliable measures of procedural biases (i.e. disparate treatment). Moreover, while procedural bias is indicative of disparate treatment, we show that the measure of procedural bias one ought to adopt is dependent on the account of the wrongfulness of disparate treatment one endorses. Finally, given that neither score-relative nor behavior-relative measures of error bias capture the relevant normative considerations, we suggest that error bias proper is best measured by score-based measures of accuracy, such as the Brier score. (shrink)

is every bit as intelligible and philosophically respectable as many other doctrines currently in favor, e.g., the doctrine that mental events are identical with brain events; the attempt to give a linguistic construal of this latter doctrine meets many of the same sorts of difficulties encountered above (see Hempel, op. cit.). Secondly, I think that evidence for universal determinism may not, as a matter of fact, be so hard to come by as one might imagine. It is a striking fact (...) about our world that we never observe any genuine cases of parallelism; it always seems possible to design some sort of interaction between any two genuine empirical magnitudes. If this is correct, then a true theory T can be deterministic only if universal determinism reigns. (shrink)

A general principle for good pedagogic strategy is this: other things equal, make the essential principles of the subject explicit rather than tacit. We think that this principle is routinely violated in conventional instruction in statistics. Even though most of the early history of probability theory has been driven by causal considerations, the terms “cause” and “causation” have practically disappeared from statistics textbooks. Statistics curricula guide students away from the concept of causality, into remembering perhaps the cliche disclaimer “correlation does (...) not mean causation,” but rarely thinking about what correlation does mean. The treatment of causality is a serious handicap to later studies of such topics as experimental design, where often the main goal is to establish (or disprove) causation. Much of the technical vocabulary of the language used in research design textbooks consists in euphemisms for specific causal relations, e.g, “latent variable,” “intervening variable,” “confounding factor,” etc. The multiplicity of terms used to refer to causation results in confusion and, in effect, may hinder understanding of the basic principles of research design. (shrink)

In her 1996 book, Error and the Growth of Experimental Knowledge, Deborah Mayo argues that use- (or heuristic) novelty is not a criterion we need to consider in assessing the evidential value of observations. Using the notion of a ''severe'' test, Mayo claims that such novelty is valuable only when it leads to severity, and never otherwise. To illustrate her view, she examines the historical case involving the famous 1919 British eclipse expeditions that generated observations supporting Einstein's theory of gravitation (...) over Newton's. My plan here is to defend use-novelty as a valuable methodological principle. I begin by exposing a weakness in Mayo's criticism of use-novelty. Remedying this weakness re-establishes the worth of use-novelty under specific conditions; in particular, heuristically novel data are to be preferred, as I will say, ''prima facie''. Armed with this revised version of use-novelty, I re-examine the history of the eclipse experiments and offer an interpretation of this episode that to an extent-and contrary to Mayo-restores the mildly heretical, Earman/Glymour evaluation of this episode offered in their (1980). I conclude by responding to criticism of my assessment of Mayo's work. (shrink)

Recent discussions of experimental tests of the Sum Rule have been carried out in the context of the special circumstances attending the Cross-Ramsey experiment. A more general analysis of possible tests is presented. A technical mistake of Fine and Glymour concerned with a misunderstanding of the physics of the Cross-Ramsey experiment is explained and a detailed analysis of a thought experiment based on the Einstein-Podolsky-Rosen wave function is given. It is concluded, in agreement with Fine, that scattering experiments do (...) not test the Sum Rule as a principle which supplements standard quantum mechanics. (shrink)

In her 1996 book, Error and the Growth of Experimental Knowledge, Deborah Mayo argues that use- (or heuristic) novelty is not a criterion we need to consider in assessing the evidential value of observations. Using the notion of a “severe” test, Mayo claims that such novelty is valuable only when it leads to severity, and never otherwise. To illustrate her view, she examines the historical case involving the famous 1919 British eclipse expeditions that generated observations supporting Einstein's theory of gravitation (...) over Newton's. My plan here is to defend use-novelty as a valuable methodological principle. I begin by exposing a weakness in Mayo's criticism of use-novelty. Remedying this weakness re-establishes the worth of use-novelty under specific conditions; in particular, heuristically novel data are to be preferred, as I will say, “prima facie”. Armed with this revised version of use-novelty, I re-examine the history of the eclipse experiments and offer an interpretation of this episode that to an extent—and contrary to Mayo—restores the mildly heretical, Earman/Glymour evaluation of this episode offered in their (1980). I conclude by responding to criticism of my assessment of Mayo's work. (shrink)

Since the introduction of mathematical population genetics, its machinery has shaped our fundamental understanding of natural selection. Selection is taken to occur when differential fitnesses produce differential rates of reproductive success, where fitnesses are understood as parameters in a population genetics model. To understand selection is to understand what these parameter values measure and how differences in them lead to frequency changes. I argue that this traditional view is mistaken. The descriptions of natural selection rendered by population genetics models are (...) in general neither predictive nor explanatory and introduce avoidable conceptual confusions. I conclude that a correct understanding of natural selection requires explicitly causal models of reproductive success. *Received May 2006; revised December 2006. †To contact the author, please write to: Department of Philosophy, Kansas State University, 201 Dickens Hall, Manhattan, KS 66506; e‐mail: glymour@ksu.edu . (shrink)

We consider the dispute between causal decision theorists and evidential decision theorists over Newcomb-like problems. We introduce a framework relating causation and directed graphs developed by Spirtes et al. (1993) and evaluate several arguments in this context. We argue that much of the debate between the two camps is misplaced; the disputes turn on the distinction between conditioning on an event E as against conditioning on an event I which is an action to bring about E. We give the essential (...) machinery for calculating the effect of an intervention and consider recent work which extends the basic account given here to the case where causal Knowledge is incomplete. (shrink)

This is a welcome reprint of a book that continues to grow in importance.

The Description for this book, Theory and Evidence, will be forthcoming.

Readable and accessible collection of papers re thinking machines.

We argue that current discussions of criteria for actual causation are ill-posed in several respects. (1) The methodology of current discussions is by induction from intuitions about an infinitesimal fraction of the possible examples and counterexamples; (2) cases with larger numbers of causes generate novel puzzles; (3) “neuron” and causal Bayes net diagrams are, as deployed in discussions of actual causation, almost always ambiguous; (4) actual causation is (intuitively) relative to an initial system state since state changes are relevant, but (...) most current accounts ignore state changes through time; (5) more generally, there is no reason to think that philosophical judgements about these sorts of cases are normative; but (6) there is a dearth of relevant psychological research that bears on whether various philosophical accounts are descriptive. Our skepticism is not directed towards the possibility of a correct account of actual causation; rather, we argue that standard methods will not lead to such an account. A different approach is required. Once upon a time a hungry wanderer came into a village. He filled an iron cauldron with water, built a fire under it, and dropped a stone into the water. “I do like a tasty stone soup” he announced. Soon a villager added a cabbage to the pot, another added some salt and others added potatoes, onions, carrots, mushrooms, and so on, until there was a meal for all. (shrink)

It is “well known” that in linear models: (1) testable constraints on the marginal distribution of observed variables distinguish certain cases in which an unobserved cause jointly influences several observed variables; (2) the technique of “instrumental variables” sometimes permits an estimation of the influence of one variable on another even when the association between the variables may be confounded by unobserved common causes; (3) the association (or conditional probability distribution of one variable given another) of two variables connected by a (...) path or pair of paths with a single common vertex (a trek) can be computed directly from the parameter values associated with each edge in the trek; (4) the association of two variables produced by multiple treks can be computed from the parameters associated with each trek; and (5) the independence of two variables conditional on a third implies the corresponding independence of the sums of the variables over all units conditional on the sums over all units of each of the original conditioning variables. (shrink)

Using Gebharter’s representation, we consider aspects of the problem of discovering the structure of unmeasured submechanisms when the variables in those submechanisms have not been measured. Exploiting an early insight of Sober’s, we provide a correct algorithm for identifying latent, endogenous structure—submechanisms—for a restricted class of structures. The algorithm can be merged with other methods for discovering causal relations among unmeasured variables, and feedback relations between measured variables and unobserved causes can sometimes be learned.

Clark Glymour, Richard Scheines, Peter Spirtes and Kevin Kelly. Discovering Causal Structure: Artifical Intelligence, Philosophy of Science and Statistical Modeling.

Reflectance spectroscopy is a standard tool for studying the mineral composition of rock and soil samples and for remote sensing of terrestrial and extraterrestrial surfaces. We describe research on automated methods of mineral identification from reflectance spectra and give evidence that a simple algorithm, adapted from a well-known search procedure for Bayes nets, identifies the most frequently occurring classes of carbonates with reliability equal to or greater than that of human experts. We compare the reliability of the procedure to the (...) reliability of several other automated methods adapted to the same purpose. Evidence is given that the procedure can be applied to some other mineral classes as well. Since the procedure is fast with low memory requirements, it is suitable for on- board scientific analysis by orbiters or surface rovers. (shrink)

Jaegwon Kim's Mind in a Physical World is an argument about mental causation that provides both a metaphysical theory and a lucid commentary on contemporary philosophical views. While I strongly recommend Kim's book to anyone interested in the subject, my endorsement is not unconditional, because I cannot make the same recomendation of the subject itself. Considering arguments of Davidson, Putnam, Burge, Block, and Kim himself, I conclude that the subject turns on a variety of implausible but received arguments, and that (...) a useful study of mental causation cannot be divorced from scientific details of cognitive psychology, physics, and neuroscience. (shrink)

One construal of convergent realism is that for each clear question, scientific inquiry eventually answers it. In this paper we adapt the techniques of formal learning theory to determine in a precise manner the circumstances under which this ideal is achievable. In particular, we define two criteria of convergence to the truth on the basis of evidence. The first, which we call EA convergence, demands that the theorist converge to the complete truth "all at once". The second, which we call (...) AE convergence, demands only that for every sentence in the theorist's language, there is a time at which the theorist settles the status of the sentence. The relative difficulties of these criteria are compared for effective and ineffective agents. We then examine in detail how the enrichment of an agent's hypothesis language makes the task of converging to the truth more difficult. In particular, we parametrize first-order languages by predicate and function symbol arity, presence or absence of identity, and quantifier prefix complexity. For nearly each choice of values of these parameters, we determine the senses in which effective and ineffective agents can converge to the complete truth on an arbitrary structure for the language. Finally, we sketch directions in which our learning theoretic setting can be generalized or made more realistic. (shrink)

Halvorson argues through a series of examples and a general result due to Myers that the “semantic view” of theories has no available account of formal theoretical equivalence. De Bouvere provides criteria overlooked in Halvorson’s paper that are immune to his counterexamples and to the theorem he cites. Those criteria accord with a modest version of the semantic view that rejects some of Van Fraassen’s apparent claims while retaining the core of Patrick Suppes’s proposal. I do not endorse any version (...) of the semantic view of theories. (shrink)

Kevin T. Kelly and Clark Glymour. Inductive Inference from Theory-Laden Data.

Your use of the JSTOR archive indicates your acceptance of J STOR’s Terms and Conditions of Use, available at http://www.jstor.org/about/terms.html. J STOR’s Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non—commercial use.

Your use of the JSTOR archive indicates your acceptance of J STOR’s Terms and Conditions of Use, available at http://www.jstor.org/about/tenns.htm1. J STOR’s Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non—commercial use.

We argue that current discussions of criteria for actual causation are ill-posed in several respects. (1) The methodology of current discussions is by induction from intuitions about an infinitesimal fraction of the possible examples and counterexamples; (2) cases with larger numbers of causes generate novel puzzles; (3) "neuron" and causal Bayes net diagrams are, as deployed in discussions of actual causation, almost always ambiguous; (4) actual causation is (intuitively) relative to an initial system state since state changes are relevant, but (...) most current accounts ignore state changes through time; (5) more generally, there is no reason to think that philosophical judgements about these sorts of cases are normative; but (6) there is a dearth of relevant psychological research that bears on whether various philosophical accounts are descriptive. Our skepticism is not directed towards the possibility of a correct account of actual causation; rather, we argue that standard methods will not lead to such an account. A different approach is required. (shrink)

Some Philosophical Prehistory of General Relativity As history, my remarks will form rather a medley. If they can claim any sort of unity (apart from a ...

In an essay recently published in this journal, Branden Fitelson argues that a variant of Miller’s argument for the language dependence of the accuracy of predictions can be applied to Joyce’s notion of accuracy of credences formulated in terms of scoring rules, resulting in a general potential problem for Joyce’s argument for probabilism. We argue that no relevant problem of the sort Fitelson supposes arises since his main theorem and his supporting arguments presuppose the validity of nonlinear transformations of credence (...) functions that Joyce’s theory, charitably construed, would identify as invalid on the basis of the principle of simple dominance. (shrink)

We agree with Cramer et al.'s goal of the discovery of causal relationships, but we argue that the authors' characterization of latent variable models (as deployed for such purposes) overlooks a wealth of extant possibilities. We provide a preliminary analysis of their data, using existing algorithms for causal inference and for the specification of latent variable models.

Various algorithms have been proposed for learning (partial) genetic regulatory networks through systematic measurements of differential expression in wild type versus strains in which expression of specific genes has been suppressed or enhanced, as well as for determining the most informative next experiment in a sequence. While the behavior of these algorithms has been investigated for toy examples, the full computational complexity of the problem has not received sufficient attention. We show that finding the true regulatory network requires (in the (...) worst-case) exponentially many experiments (in the number of genes). Perhaps more importantly, we provide an algorithm for determining the set of regulatory networks consistent with the observed data. We then show that this algorithm is infeasible for realistic data (specifically, nine genes and ten experiments). This infeasibility is not due to an algorithmic flaw, but rather to the fact that there are far too many networks consistent with the data (10 18 in the provided example). We conclude that gene perturbation experiments are useful in confirming regulatory network models discovered by other techniques, but not a feasible search strategy. (shrink)

Recent literature in philosophy of science has addressed purported notions of explanatory virtues—‘explanatory power’, ‘unification’, and ‘coherence’. In each case, a probabilistic relation between a theory and data is said to measure the power of an explanation, or degree of unification, or degree of coherence. This essay argues that the measures do not capture cases that are paradigms of scientific explanation, that the available psychological evidence indicates that the measures do not capture judgements of explanatory power, and, finally, that the (...) measures do not provide useful methods for selecting hypotheses. 1. Introduction2. Some Proposed Measures of Explanatory Virtues3. Descriptive Inadequacy3.1 Excellent but false explanations3.2 Causal explanation4. Psychological Inadequacy5. Finding the Truth6. Conclusion. (shrink)