Elliott Sober

This paper defends two theses about probabilistic reasoning. First, although modus ponens has a probabilistic analog, modus tollens does not – the fact that a hypothesis says that an observation is very improbable does not entail that the hypothesis is improbable. Second, the evidence relation is essentially comparative; with respect to hypotheses that confer probabilities on observation statements but do not entail them, an observation O may favor one hypothesis H1 over another hypothesis H2 , but O cannot be said to confirm or disconfirm H1 without such relativization. These points have serious consequences for the Intelligent Design movement. Even if evolutionary theory entailed that various complex adaptations are very improbable, that would neither disconfirm the theory nor support the hypothesis of intelligent design. For either of these conclusions to follow, an additional question must be answered: With respect to the adaptive features that evolutionary theory allegedly says are very improbable, what is their probability of arising if they were produced by intelligent design? This crucial question has not been addressed by the ID movement.

Akaike’s framework for thinking about model selection in terms of the goal of predictive accuracy and his criterion for model selection have important philosophical implications. Scientists often test models whose truth values they already know, and they often decline to reject models that they know full well are false. Instrumentalism helps explain this pervasive feature of scientific practice, and Akaike’s framework helps provide instrumentalism with the epistemology it needs. Akaike’s criterion for model selection also throws light on the role of parsimony considerations in hypothesis evaluation. I explain the basic ideas behind Akaike’s framework and criterion; several biological examples, including the use of maximum likelihood methods in phylogenetic inference, are considered.

The chapter discusses the principle of conservatism and traces how the general principle is related to the specific one. This tracing suggests that the principle of conservatism needs to be refined. Connecting the principle in cognitive science to more general questions about scientific inference also allows us to revisit the question of realism versus instrumentalism. The framework deployed in model selection theory is very general; it is not specific to the subject matter of science. The chapter outlines some non-Bayesian ideas that have been developed in model selection theory. The principle of conservatism, like C. Lloyd Morgan's canon, describes a preference concerning kinds of parameters. It says that a model that postulates only lower-level intentionality is preferable to one that postulates higher-level intentionality if both fit the data equally well. The model selection approach to parsimony helps explain why unification is a theoretical virtue.

an observation to formulate a theory, it is no surprise that the resulting theory accurately captures that observation. However, when the theory makes a novel prediction—when it predicts an observation that was not used in its formulation—this seems to provide more substantial confirmation of the theory. This paper presents a new approach to the vexed problem of understanding the epistemic difference between prediction and accommodation. In fact, there are several problems that need to be disentangled; in all of them, the key is the concept of overfitting. We float the hypothesis that accommodation is a defective methodology only when the methods used to accommodate the data fail to guard against the risk of overfitting. We connect our analysis with the proposals that other philosophers have made. We also discuss its bearing on the conflict between instrumentalism and scientific realism. Introduction Predictivisms—a taxonomy Observations Formulating the problem What might Annie be doing wrong? Solutions Observations explained Mayo on severe tests The miracle argument and scientific realism Concluding comments.

We argue in Roche and Sober (2013) that explanatoriness is evidentially irrelevant in that Pr(H | O&EXPL) = Pr(H | O), where H is a hypothesis, O is an observation, and EXPL is the proposition that if H and O were true, then H would explain O. This is a “screening-off” thesis. Here we clarify that thesis, reply to criticisms advanced by Lange (2017), consider alternative formulations of Inference to the Best Explanation, discuss a strengthened screening-off thesis, and consider how it bears on the claim that unification is evidentially relevant.

As every philosopher knows, “the design argument” concludes that God exists from premisses that cite the adaptive complexity of organisms or the lawfulness and orderliness of the whole universe. Since 1859, it has formed the intellectual heart of creationist opposition to the Darwinian hypothesis that organisms evolved their adaptive features by the mindless process of natural selection. Although the design argument developed as a defense of theism, the logic of the argument in fact encompasses a larger set of issues. William Paley saw clearly that we sometimes have an excellent reason to postulate the existence of an intelligent designer. If we find a watch on the heath, we reasonably infer that it was produced by an intelligent watchmaker. This design argument makes perfect sense. Why is it any different to claim that the eye was produced by an intelligent designer? Both critics and defenders of the design argument need to understand what the ground rules are for inferring that an intelligent designer is the unseen cause of an observed effect.

All extant purely probabilistic measures of explanatory power satisfy the following technical condition: if Pr(E | H1) > Pr(E | H2) and Pr(E | ∼H1) < Pr(E | ∼H2), then H1’s explanatory power with respect to E is greater than H2’s explanatory power with respect to E. We argue that any measure satisfying this condition faces three serious problems—the Problem of Temporal Shallowness, the Problem of Negative Causal Interactions, and the Problem of Nonexplanations. We further argue that many such measures face a fourth problem—the Problem of Explanatory Irrelevance.

Solipsism poses a familiar epistemological problem. Each of us has beliefs about a world that allegedly exists outside our own minds. The problem is to justify these nonsolipsistic convictions. One standard approach is to argue that the existence of things outside our own sensations may reasonably be inferred from regularities that obtain within our sensations. Certain experiences, which I will call tiger sounds and tiger visual images, exhibit a striking correlation. We can explain the existence of this correlation by postulating an entity that is a cause of both. If there were tigers, it would be no surprise that certain sights and certain sounds tend to co-occur. Our rejection of solipsism can thus be justified by appeal to an abductive argument; we advance an inference to the best explanation conforming to the pattern that Reichenbach called the principle of the common cause.

We begin by considering two principles, each having the form causal completeness ergo screening-off. The first concerns a common cause of two or more effects; the second describes an intermediate link in a causal chain. They are logically independent of each other, each is independent of Reichenbach's principle of the common cause, and each is a consequence of the causal Markov condition. Simple examples show that causal incompleteness means that screening-off may fail to obtain. We derive a stronger result: in a rather general setting, if the composite cause C1 & C2 & … & Cn screens-off one event from another, then each of the n component causes C1, C2, …, Cn must fail to screen-off. The idea that a cause may be ordinally invariant in its impact on different effects is defined; it plays an important role in establishing this no-go theorem. Along the way, we describe how composite and component causes can all screen-off when ordinal invariance fails. We argue that this theorem is relevant to assessing the plausibility of the two screening-off principles. The discovery of incomplete causes that screen-off is not evidence that causal completeness must engender screening-off. Formal and epistemic analogies between screening-off and determinism are discussed.

Human beings are peculiar. In laboratory experiments, they often cooperate in one-shot prisoners’ dilemmas, they frequently offer 1/2 and reject low offers in the ultimatum game, and they often bid 1/2 in the game of divide-the-cake All these behaviors are puzzling from the point of view of game theory. The first two are irrational, if utility is measured in a certain way.1 The last isn’t positively irrational, but it is no more rational than other possible actions, since there are infinitely many other Nash equilibria besides the one in which both players bid 1/2. At the same time, these behaviors seem to indicate that people are sometimes inclined to be cooperative, fair, and just. In his stimulating new book, Brian Skyrms sets himself the task of showing why these inclinations evolved, or how they might have evolved, under the pressure of natural selection. The goal is not to justify our ethical intuitions, but to explain why we have them.2..

Decision theory requires agents to assign probabilities to states of the world and utilities to the possible outcomes of different actions. When agents commit to having the probabilities and/or utilities in a decision problem defined by objective features of the world, they may find themselves unable to decide which actions maximize expected utility. Decision theory has long recognized that work-around strategies are available in special cases; this is where dominance reasoning, minimax, and maximin play a role. Here we describe a different work around, wherein a rational decision about one decision problem can be reached by ‘interpolating’ information from another problem that the agent believes has already been rationally solved.

Only one traditional objection to Pascal's wager is telling: Pascal assumes a particular theology, but without justification. We produce two new objections that go deeper. We show that even if Pascal's theology is assumed to be probable, Pascal's argument does not go through. In addition, we describe a wager that Pascal never considered, which leads away from Pascal's conclusion. We then consider the impact of these considerations on other prudential arguments concerning what one should believe, and on the more general question of when and why belief formation ought to be based solely on the evidence.

The hypothesis of group selection fell victim to a seemingly devastating critique in 1960s evolutionary biology. In Unto Others (1998), we argue to the contrary, that group selection is a conceptually coherent and empirically well documented cause of evolution. We suggest, in addition, that it has been especially important in human evolution. In the second part of Unto Others, we consider the issue of psychological egoism and altruism -- do human beings have ultimate motives concerning the well-being of others? We argue that previous psychological and philosophical work on this question has been inconclusive. We propose an evolutionary argument for the claim that human beings have altruistic ultimate motives.

A model of a Markov process is presented in which observing the present state of a system is asymmetrically related to inferring the system's future and inferring its past. A likelihood inference about the system's past state, based on observing its present state, is justified no matter what the parameter values in the model happen to be. In contrast, a probability inference of the system's future state, based on observing its present state, requires further information about the parameter values

Nancy Cartwright (1983, 1999) argues that (1) the fundamental laws of physics are true when and only when appropriate ceteris paribus modifiers are attached and that (2) ceteris paribus modifiers describe conditions that are almost never satisfied. She concludes that when the fundamental laws of physics are true, they don't apply in the real world, but only in highly idealized counterfactual situations. In this paper, we argue that (1) and (2) together with an assumption about contraposition entail the opposite conclusion — that the fundamental laws of physics do apply in the real world. Cartwright extracts from her thesis about the inapplicability of fundamental laws the conclusion that they cannot figure in covering-law explanations. We construct a different argument for a related conclusion — that forward-directed idealized dynamical laws cannot provide covering-law explanations that are causal. This argument is neutral on whether the assumption about contraposition is true. We then discuss Cartwright's simulacrum account of explanation, which seeks to describe how idealized laws can be explanatory.

We argue against some of Reichenbach's claims about causal forks are incorrect. We do not see why the Second Law of Thermodynamics rules out the existence of conjunctive forks open to the past. In addition, we argue that a common effect rarely forms a conjunctive fork with its joint causes, but it sometimes does. Nevertheless, we think there is something to be said for Reichenbach's idea that forks of various kinds are relevant to explaining why we know more about the past than about the future. And entropy turns out to play a role in the explanation, though not the determining role that Reichenbach envisions.

We learned from Good that there is no saying whether a black raven confirms the generalization that all ravens are black unless one is prepared to make substantive background assumptions. The same point, applied to the problem of other minds, is that the mere observation that Self and Other share certain behaviors and that Self has a mind is not enough. The problem of other minds turns into the problem of searching out common causes. This paper presents a probabilistic representation of the problem of other minds, and shows that the likelihood concept provides a framework for understanding the Self and Other problem.

In my paper “Intelligent Design Theory and the Supernatural—the ‘God or Extra-Terrestrial’ Reply,” I argued that Intelligent Design (ID) Theory, when coupled with independently plausible further assumptions, leads to the conclusion that a supernatural intelligent designer exists. ID theory is therefore not neutral on the question of whether there are supernatural agents. In this respect, it differs from the Darwinian theory of evolution. John Beaudoin replies to my paper in his “Sober on Intelligent Design Theory and the Intelligent Designer,” arguing that my paper faces two challenges. In the present paper, I try to address Beaudoin’s challenges.

How could the fundamental mental operations which facilitate scientific theorizing be the product of natural selection, since it appears that such theoretical methods were neither used nor useful "in the cave"-i.e., in the sequence of environments in which selection took place? And if these wired-in information processing techniques were not selected for, how can we view rationality as an adaptation? It will be the purpose of this paper to address such questions as these, and in the process to sketch some of the considerations that an evolutionary account of rationality may involve. By describing the broad framework within which the evolution of rationality may eventually be understood, I hope to undermine the idea that evolutionary theory is somehow incapable of dealing with this characteristic and requires supplementation by some novel principle. A more modest ambition of the paper is to try to provoke those who think that there are special problems involved in this evolutionary inquiry to say what these problems are.

In both biology and the human sciences, social groups are sometimes treated as adaptive units whose organization cannot be reduced to individual interactions. This group-level view is opposed by a more individualistic one that treats social organization as a byproduct of self-interest. According to biologists, group-level adaptations can evolve only by a process of natural selection at the group level. Most biologists rejected group selection as an important evolutionary force during the 1960s and 1970s but a positive literature began to grow during the 1970s and is rapidly expanding today. We review this recent literature and its implications for human evolutionary biology. We show that the rejection of group selection was based on a misplaced emphasis on genes as “replicators” which is in fact irrelevant to the question of whether groups can be like individuals in their functional organization. The fundamental question is whether social groups and other higher-level entities can be “vehicles” of selection. When this elementary fact is recognized, group selection emerges as an important force in nature and what seem to be competing theories, such as kin selection and reciprocity, reappear as special cases of group selection. The result is a unified theory of natural selection that operates on a nested hierarchy of units.The vehicle-based theory makes it clear that group selection is an important force to consider in human evolution. Humans can facultatively span the full range from self-interested individuals to “organs” of group-level “organisms.” Human behavior not only reflects the balance between levels of selection but it can also alter the balance through the construction of social structures that have the effect of reducing fitness differences within groups, concentrating natural selection (and functional organization) at the group level. These social structures and the cognitive abilities that produce them allow group selection to be important even among large groups of unrelated individuals.