A cause may influence its effect via multiple paths. Paradigmatically (Hesslow [1974]), taking birth control pills both decreases one’s risk of thrombosis by preventing pregnancy and increases it by producing a blood chemical. Building on Pearl ([2001]), I explicate the notion of a path-specific effect. Roughly, a path-specific effect of C on E via path P is the degree to which a change in C would change E were they to be transmitted only via P. Facts about such effects may (...) be gleaned from the structural equations commonly used to represent the causal relationships among variables. I contrast my analysis of the Hesslow case with those given by theorists of probabilistic causality, who mistakenly link it to issues of causal heterogeneity, token-causation and indeterminism. The reason probabilistic theories misdiagnose this case is that they pay inadequate attention to the structural relationships among variables. (shrink)

Within the causal modeling literature, debates about the Causal Faithfulness Condition have concerned whether it is probable that the parameters in causal models will have values such that distinct causal paths will cancel. As the parameters in a model are fixed by the probability distribution over its variables, it is initially puzzling what it means to assign probabilities to these parameters. I propose that to assign a probability to a parameter in a model is to treat that parameter as a (...) function of a variable in an augmented model. By combining this proposal with widely adopted principles regarding which variables must be included in a model, I argue that the various proposed counterexamples to CFC involving coordinated parameters are not genuine counterexamples. I then consider the cases in which CFC fails due not to coordination, but by coincidence, and propose explanatory and predictive bases for ruling out such coincidences without presuming that they are improbable. The aim of the proposed defenses is not to show that CFC never fails, but rather to argue that its use in a particular context may be defended using general modeling assumptions rather than by relying on claims about how often it fails. (shrink)

Many generic sentences express stable inductive generalizations. Stable inductive generalizations are typically true for a causal reason. In this paper we investigate to what extent this is also the case for the generalizations expressed by generic sentences. More in particular, we discuss the possibility that many generic sentences of the form ‘ks have feature e’ are true because kind k have the causal power to ‘produce’ feature e. We will argue that such an analysis is quite close to a probabilistic (...) based analysis of generic sentences according to which ‘relatively many’ ks have feature e, and that, in fact, this latter type of analysis can be ‘grounded’ in terms of causal powers. We will argue, moreover, that the causal power analysis is sometimes preferred to a correlation-based analysis, because it takes the causal structure that gives rise to the probabilistic data into account. (shrink)

Inductive reasoning requires exploiting links between evidence and hypotheses. This can be done focusing either on the posterior probability of the hypothesis when updated on the new evidence or on the impact of the new evidence on the credibility of the hypothesis. But are these two cognitive representations equally reliable? This study investigates this question by comparing probability and impact judgments on the same experimental materials. The results indicate that impact judgments are more consistent in time and more accurate than (...) probability judgments. Impact judgments also predict the direction of errors in probability judgments. These findings suggest that human inductive reasoning relies more on estimating evidential impact than on posterior probability. (shrink)

Data from medical research are typically summarized with various types of outcome measures. We present three arguments in favor of absolute over relative outcome measures. The first argument is from cognitive bias: relative measures promote the reference class fallacy and the overestimation of treatment effectiveness. The second argument is decision-theoretic: absolute measures are superior to relative measures for making a decision between interventions. The third argument is causal: interpreted as measures of causal strength, absolute measures satisfy a set of desirable (...) properties, but relative measures do not. Absolute outcome measures outperform relative measures on all counts. (shrink)

This paper develops axiomatic foundations for a probabilistic-interventionist theory of causal strength. Transferring methods from Bayesian confirmation theory, I proceed in three steps: I develop a framework for defining and comparing measures of causal strength; I argue that no single measure can satisfy all natural constraints; I prove two representation theorems for popular measures of causal strength: Pearl's causal effect measure and Eells' difference measure. In other words, I demonstrate these two measures can be derived from a set of plausible (...) adequacy conditions. The paper concludes by sketching future research avenues. (shrink)

The propensity interpretation of fitness draws on the propensity interpretation of probability, but advocates of the former have not attended sufficiently to problems with the latter. The causal power of C to bring about E is not well-represented by the conditional probability Pr. Since the viability fitness of trait T is the conditional probability Pr, the viability fitness of the trait does not represent the degree to which having the trait causally promotes surviving. The same point holds for fertility fitness. (...) This failure of trait fitness to capture causal role can also be seen in the fact that coextensive traits must have the same fitness values even if one of them promotes survival and the other is neutral or deleterious. Although the fitness of a trait does not represent the trait’s causal power to promote survival and reproduction, variation in fitness in a population causally promotes change in trait frequencies; in this sense, fitness variation is a population-level propensity. (shrink)

Carl Hempel argued that probabilistic hypotheses are limited in what they can explain. He contended that a hypothesis cannot explain why E is true if the hypothesis says that E has a probability less than 0.5. Wesley Salmon and Richard Jeffrey argued to the contrary, contending that P can explain why E is true even when P says that E’s probability is very low. This debate concerned noncontrastive explananda. Here, a view of contrastive causal explanation is described and defended. It (...) provides a new limit on what probabilistic hypotheses can explain; the limitation is that P cannot explain why E is true rather than A if P assign E a probability that is less than or equal to the probability that P assigns to A. The view entails that a true deterministic theory and a true probabilistic theory that apply to the same explanandum partition are such that the deterministic theory explains all the true contrastive propositions constructable from that partition, whereas the probabilistic theory often fails to do so. (shrink)

Two recently popular metaphilosophical movements, formal philosophy and experimental philosophy, promote what seem to be conflicting methodologies. Nonetheless, I argue that the two can be mutually supportive. I propose an experimentally-informed variation on explication, a powerful formal philosophical tool introduced by Carnap. The resulting method, which I call “experimental explication,” provides the formalist with a means of responding to explication's gravest criticism. Moreover, this method introduces a philosophically salient, positive role for survey-style experiments while steering clear of several objections that (...) critics of “positive experimental philosophy” raise. Thus, it provides the experimentalist with a more defensible example of how empirical work can have positive philosophical import. For these reasons, experimental explication should appeal to experimental philosophers and formal philosophers alike. (shrink)

What is a valid measuring instrument? Recent philosophy has attended to logic of justification of measures, such as construct validation, but not to the question of what it means for an instrument to be a valid measure of a construct. A prominent approach grounds validity in the existence of a causal link between the attribute and its detectable manifestations. Some of its proponents claim that, therefore, validity does not depend on pragmatics and research context. In this paper, I cast doubt (...) on the possibility of a context-independent causal account of validity. I assess several versions, arguing that all of them fail to judge the validity of measuring instruments correctly. Because different research purposes require different properties from measuring instruments, no account of validity succeeds without referring to the specific research purpose that creates the need for measurement in the first place. (shrink)

This paper presents a symmetry condition for probabilistic measures of confirmation which is weaker than commutativity symmetry, disconfirmation commutativity symmetry but also antisymmetry. It is based on the idea that for any value a probabilistic measure of confirmation can assign there is a corresponding case where degrees of confirmation are symmetric. It is shown that a number of prominent confirmation measures such as Carnap’s difference function, Rescher’s measure of confirmation, Gaifman’s confirmation rate and Mortimer’s inverted difference function do not satisfy (...) this condition and instead exhibit a previously unnoticed and rather puzzling behavior in certain cases of disconfirmation. This behavior also carries over to probabilistic measures of information change, causal strength, explanatory power and coherence. (shrink)

It is often natural to compare two events by describing one as ‘more of a cause’ of some effect than the other. But what do such comparisons amount to, exactly? This paper aims to provide a guided tour of the recent literature on ‘degrees of causation’. Section 2 looks at what I call ‘dependence measures’, which arise from thinking of causes as difference‐makers. Section 3 looks at what I call ‘production measures’, which arise from thinking of causes as jointly sufficient (...) for their effects. Finally, section 4 examines the important question of whether there is any sense in which an agent is more responsible for an outcome in virtue of her action being more of a cause of it. I describe a puzzle that emerges from this question, and explore various strategies for resolving it. (shrink)

Existing research suggests that people's judgments of actual causation can be influenced by the degree to which they regard certain events as normal. We develop an explanation for this phenomenon that draws on standard tools from the literature on graphical causal models and, in particular, on the idea of probabilistic sampling. Using these tools, we propose a new measure of actual causal strength. This measure accurately captures three effects of normality on causal judgment that have been observed in existing studies. (...) More importantly, the measure predicts a new effect ("abnormal deflation"). Two studies show that people's judgments do, in fact, show this new effect. Taken together, the patterns of people's causal judgments thereby provide support for the proposed explanation. (shrink)

How do people judge the degree of causal responsibility that an agent has for the outcomes of her actions? We show that a relatively unexplored factor -- the robustness of the causal chain linking the agent’s action and the outcome -- influences judgments of causal responsibility of the agent. In three experiments, we vary robustness by manipulating the number of background circumstances under which the action causes the effect, and find that causal responsibility judgments increase with robustness. In the first (...) experiment, the robustness manipulation also raises the probability of the effect given the action. Experiments 2 and 3 control for probability-raising, and show that robustness still affects judgments of causal responsibility. In particular, Experiment 3 introduces an Ellsberg type of scenario to manipulate robustness, while keeping the conditional probability and the skill deployed in the action fixed. Experiment 4, replicates the results of Experiment 3, while contrasting between judgments of causal strength and of causal responsibility. The results show that in all cases, the perceived degree of responsibility increases with the robustness of the action-outcome causal chain. (shrink)

Which country, politician, or policy is more of a cause of the Covid-19 pandemic death toll? Which of the two factories causally contributed more to the pollution of the nearby river? A wide-ranging portion of our everyday thought and talk, and attitudes rely on a graded notion of causation. However, it is sometimes highlighted that on most contemporary accounts, causation is on-off. Some philosophers further question the legitimacy of talk of degrees of causation and suggest that we avoid it. Some (...) hold that the notion of degrees of causation is an illusion. In this paper, I’ll argue that causation does come in degrees. (shrink)

Several prominent debates in biology, such as those surrounding adaptationism, group selection, and punctuated equilibrium, have focused on disagreements about the relative importance of a cause in producing a phenomenon of interest. Some philosophers, such as John Beatty have expressed scepticism about the scientific value of engaging in these controversies, and Karen Kovaka has suggested that their value might be limited. In this paper, I challenge that scepticism by giving a novel analysis of relative significance controversies, showing that there are (...) three forms they can take. I argue that these controversies can have significant epistemic upshots, in that they help scientists form predictions about new instances of the phenomenon of interest. Finally, using two historical examples, I show how engaging in these controversies can improve our understanding of causal relationships. (shrink)

Probability ratio and likelihood ratio measures of inductive support and related notions have appeared as theoretical tools for probabilistic approaches in the philosophy of science, the psychology of reasoning, and artificial intelligence. In an effort of conceptual clarification, several authors have pursued axiomatic foundations for these two families of measures. Such results have been criticized, however, as relying on unduly demanding or poorly motivated mathematical assumptions. We provide two novel theorems showing that probability ratio and likelihood ratio measures can be (...) axiomatized in a way that overcomes these difficulties. (shrink)

Searching for information is critical in many situations. In medicine, for instance, careful choice of a diagnostic test can help narrow down the range of plausible diseases that the patient might have. In a probabilistic framework, test selection is often modeled by assuming that people's goal is to reduce uncertainty about possible states of the world. In cognitive science, psychology, and medical decision making, Shannon entropy is the most prominent and most widely used model to formalize probabilistic uncertainty and the (...) reduction thereof. However, a variety of alternative entropy metrics (Hartley, Quadratic, Tsallis, Rényi, and more) are popular in the social and the natural sciences, computer science, and philosophy of science. Particular entropy measures have been predominant in particular research areas, and it is often an open issue whether these divergences emerge from different theoretical and practical goals or are merely due to historical accident. Cutting across disciplinary boundaries, we show that several entropy and entropy reduction measures arise as special cases in a unified formalism, the Sharma–Mittal framework. Using mathematical results, computer simulations, and analyses of published behavioral data, we discuss four key questions: How do various entropy models relate to each other? What insights can be obtained by considering diverse entropy models within a unified framework? What is the psychological plausibility of different entropy models? What new questions and insights for research on human information acquisition follow? Our work provides several new pathways for theoretical and empirical research, reconciling apparently conflicting approaches and empirical findings within a comprehensive and unified information-theoretic formalism. (shrink)

The probability that the fitter of two alleles will increase in frequency in a population goes up as the product of N (the effective population size) and s (the selection coefficient) increases. Discovering the distribution of values for this product across different alleles in different populations is a very important biological task. However, biologists often use the product Ns to define a different concept; they say that drift “dominates” selection or that drift is “stronger than” selection when Ns is much (...) smaller than some threshold quantity (e.g., ½) and that the reverse is true when Ns is much larger than that threshold. We argue that the question of whether drift dominates selection for a single allele in a single population makes no sense. Selection and drift are causes of evolution, but there is no fact of the matter as to which cause is stronger in the evolution of any given allele. (shrink)

I argue for responsibility internalism. That is, moral responsibility (i.e., accountability, or being apt for praise or blame) depends only on factors internal to agents. Employing this view, I also argue that no one is responsible for what AI does but this isn’t morally problematic in a way that counts against developing or using AI. Responsibility is grounded in three potential conditions: the control (or freedom) condition, the epistemic (or awareness) condition, and the causal responsibility condition (or consequences). I argue (...) that causal responsibility is irrelevant for moral responsibility, and that the control condition and the epistemic condition depend only on factors internal to agents. Moreover, since what AI does is at best a consequence of our actions, and the consequences of our actions are irrelevant to our responsibility, no one is responsible for what AI does. That is, the so-called responsibility gap exits. However, this isn’t morally worrisome for developing or using AI. Firstly, I argue, current AI doesn’t generate a new kind of concern about responsibility that the older technologies don’t. Then, I argue that responsibility gap is not worrisome because neither responsibility gap, nor my argument for its existence, entails that no one can be justly punished, held accountable, or incurs duties in reparations when AI causes a harm. (shrink)