Causal Reasoning, Misc

In Nozick’s rendition of the decision situation given in Newcomb’s Paradox dominance and the principle of maximum expected utility recommend different strategies. While evidential decision theory seems to be split over which principle to apply and how to interpret the principles in the first place, causal decision theory seems to go for the solution recommended by dominance. As a reply to the CDT proposal by Wolfgang Spohn, who opts for “one-boxing” by employing reflexive decision graphs, I will draw on the (...) framework of causal knowledge patterns, i.e., Bayes net causal models, augmented by non-causal knowledge, to finally arrive at “one-boxing” – more intuitively and more closely to what actually is in Nozick’s story. This proposal allows the careful re-examination of all relevant concepts in the original story and might cast new light on the following questions: How may causality in general be understood to allow causal inference from hybrid patterns encoding subjective knowledge? How can the notion of prediction be analyzed – philosophically and formally? And what’s the decision-maker’s conceptualization of the situation he will act upon? (shrink)

Research in the psychology of causal thinking has frequently revealed effects of normative considerations on causal attributions, where participants tend to assign causality more strongly to agents who violate a norm in bringing about an outcome. Across several experiments, we show that it is possible to reverse this norm effect when the outcome in question is good rather than bad: in these cases, participants assign causality more strongly to a norm-conforming agent than to an agent who violates a norm. We (...) argue that this supports an explanation of the norm effect according to which it is due to a tendency to interpret statements of the form “X caused Y” or “X is the cause of Y” as assigning responsibility to X for bringing Y about. (shrink)

There is, among some scientists and philosophers, the idea that any theory that would allow the time travel would introduce causal issues. These types of temporal paradoxes can be avoided by the Novikov self-consistency principle or by a variation in the interpretation of many worlds with interacting worlds. The world in which we live has, according to David Lewis, a Parmenidean ontology: "a manifold of events in four dimensions," and the occupants of the world are the 4-dimensional aggregates of the (...) stages - "temporal lines". The causal loops in backwards time travel involve events that appear to "come from nowhere," paradoxical "self-existent" objects or information, resulting in a bootstrap paradox. Many believe that causality loops are not impossible or unacceptable, but only inexplicable. DOI: 10.13140/RG.2.2.28792.70407. (shrink)

To date, hundreds of researchers have employed the method of Qualitative Comparative Analysis (QCA) for the purpose of causal inference. In a recent series of simulation studies, however, several authors have questioned the correctness of QCA in this connection. Some prominent representatives of the method have replied in turn that simulations with artificial data are unsuited for assessing QCA. We take issue with either position in this impasse. On the one hand, we argue that data-driven evaluations of the correctness of (...) a procedure of causal inference require artificial data. On the other hand, we prove all previous attempts in this direction to have been defective. For the first time in the literature on configurational comparative methods, we lay out a set of formal criteria for an adequate evaluation of QCA before implementing a battery of inverse-search trials to test how this method performs in different recovery contexts according to these criteria. While our results indicate that QCA is correct when generating the parsimonious solution type, they also demonstrate that the method is incorrect when generating the conservative and intermediate solution type. In consequence, researchers using QCA for causal inference, particularly in human-sensitive areas such as public health and medicine, should immediately discontinue employing the method’s conservative and intermediate search strategies. (shrink)

This paper considers an important recent contribution by Miguel Hernán to the ongoing debate about causal inference in epidemiology. Hernán rejects the idea that there is an in-principle epistemic distinction between the results of randomized controlled trials and observational studies: both produce associations which we may be more or less confident interpreting as causal. However, Hernán maintains that trials have a semantic advantage. Observational studies that seek to estimate causal effect risk issuing meaningless statements instead. The POA proposes a solution (...) to this problem: improved restrictions on the meaningful use of causal language, in particular “causal effect”. This paper argues that new restrictions in fact fail their own standards of meaningfulness. The paper portrays the desire for a restrictive definition of causal language as positivistic, and argues that contemporary epidemiology should be more realistic in its approach to causation. In a realist context, restrictions on meaningfulness based on precision of definition are neither helpful nor necessary. Hernán’s favoured approach to causal language is saved from meaninglessness, along with the approaches he rejects. (shrink)

In this paper, I critically evaluate several related, provocative claims made by proponents of data-intensive science and “Big Data” which bear on scientific methodology, especially the claim that scientists will soon no longer have any use for familiar concepts like causation and explanation. After introducing the issue, in section 2, I elaborate on the alleged changes to scientific method that feature prominently in discussions of Big Data. In section 3, I argue that these methodological claims are in tension with a (...) prominent account of scientific method, often called “Inference to the Best Explanation” (IBE). Later on, in section 3, I consider an argument against IBE that will be congenial to proponents of Big Data, namely the argument due to Roche and Sober (2013) that “explanatoriness is evidentially irrelevant”. This argument is based on Bayesianism, one of the most prominent general accounts of theory-confirmation. In section 4, I consider some extant responses to this argument, especially that of Climenhaga (2017). In section 5, I argue that Roche and Sober’s argument does not show that explanatory reasoning is dispensable. In section 6, I argue that there is good reason to think explanatory reasoning will continue to prove indispensable in scientific practice. Drawing on Cicero’s oft-neglected De Divinatione, I formulate what I call the “Ciceronian Causal-nomological Requirement”, (CCR), which states roughly that causal-nomological knowledge is essential for relying on correlations in predictive inference. I defend a version of the CCR by appealing to the challenge of “spurious correlations”, chance correlations which we should not rely upon for predictive inference. In section 7, I offer some concluding remarks. (shrink)

Causal reasoning is an aspect of learning, reasoning, and decision-making that involves the cognitive ability to discover relationships between causal relata, learn and understand these causal relationships, and make use of this causal knowledge in prediction, explanation, decision-making, and reasoning in terms of counterfactuals. Can we fully automate causal reasoning? One might feel inclined, on the basis of certain groundbreaking advances in causal epistemology, to reply in the affirmative. The aim of this paper is to demonstrate that one still has (...) good skeptical grounds for resisting any conclusions in favour of the automation of causal reasoning. If by causal reasoning is meant the entirety of the process through which we discover causal relationships and make use of this knowledge in prediction, explanation, decision-making, and reasoning in terms of counterfactuals, then one relies besides on tacit knowledge, as might be constituted by or derived from the epistemic faculty virtues and abilities of the causal reasoner, the value systems and character traits of the causal reasoner, the implicit knowledge base available to the causal reasoner, and the habits that sustain our causal reasoning practices. While certain aspects of causal reasoning may be axiomatized and formalized and algorithms may be implemented to approximate causal reasoning, one has to remain skeptical about whether causal reasoning may be fully automated. This demonstration will involve an engagement with Meno’s Paradox. (shrink)

A good definition of process tracing should highlight what is distinctive about process tracing as a methodology of causal inference. I look at eight criteria that are used to define process tracing in the methodological literature, and I dismiss all eight criteria as unhelpful (some because they are too restrictive, and others because they are vacuous). In place of these criteria, I propose four alternative criteria, and I draw a distinction between process tracing for the ultimate aim of testing a (...) start--end hypothesis versus process tracing as an ultimate end in itself. Although it is clear enough how the former method works, there is still much methodological work to be done in understanding the latter method as a distinctive method of causal inference, I argue. (shrink)

Bradford Hill (1965) highlighted nine aspects of the complex evidential situation a medical researcher faces when determining whether a causal relation exists between a disease and various conditions associated with it. These aspects are widely cited in the literature on epidemiological inference as justifying an inference to a causal claim, but the epistemological basis of the Hill aspects is not understood. We offer an explanatory coherentist interpretation, explicated by Thagard's ECHO model of explanatory coherence. The ECHO model captures the complexity (...) of epidemiological inference and provides a tractable model for inferring disease causation. We apply this model to three cases: the inference of a causal connection between the Zika virus and birth defects, the classic inference that smoking causes cancer, and John Snow’s inference about the cause of cholera. (shrink)

In cases where many causes together bring about an effect, it is common to select some as particularly important. Philosophers since Mill have been pessimistic about analyzing this reasoning due its variability and the multifarious causal and pragmatic details of how it works. I argue Mill was right to think these details matter, but wrong that they preclude philosophical analysis of causal selection. I show that analyzing the pragmatic details of scientific debates about the important causes of the Bhopal Gas (...) Tragedy can illuminate causal reasoning about disasters, and shed new light on causality and causal selection. (shrink)

This chapter explores the idea that the need to establish common knowledge is one feature that makes social cognition stand apart in important ways from cognition in general. We develop this idea on the background of the claim that social cognition is nothing but a type of causal inference. We focus on autism as our test-case, and propose that a specific type of problem with common knowledge processing is implicated in challenges to social cognition in autism spectrum disorder (ASD). This (...) problem has to do with the individual’s assessment of the reliability of messages that are passed between people as common knowledge emerges. The proposal is developed on the background of our own empirical studies and outlines different ways common knowledge might be comprised. We discuss what these issues may tell us about ASD, about the relation between social and non-social cognition, about social objects, and about the dynamics of social networks. (shrink)

This article is an introduction to the Synthese Special Issue, Philosophy of Epidemiology. The overall goals of the issue are to revisit the state of philosophy of epidemiology and to provide a forum for new voices, approaches, and perspectives in the philosophy of epidemiology literature. The introduction begins by drawing on Geoffrey Rose’s work on how to conceptualize and design interventions for populations, rather than individuals. It then goes on to highlight some themes that emerged in the articles that make (...) up the issue: philosophy of epidemiology and epidemiological theory—what they are and what they ought to be, pluralism in measurement and causal attribution, epistemic and non-epistemic values in disputes epidemiological practices, decentering philosophy of epidemiology’s Eurocentrism, letting pragmatism guide uses of big data in epidemiology, and revisiting the lessons of classic texts in epidemiological causal inference. The introduction concludes with comments on a philosophy of epidemiology debate we see on, regarding the politics of philosophy of epidemiology. (shrink)

Temporal binding refers to a phenomenon whereby the time interval between a cause and its effect is perceived as shorter than the same interval separating two unrelated events. We examined the developmental profile of this phenomenon by comparing the performance of groups of children (aged 6-7-, 7-8-, and 9-10- years) and adults on a novel interval estimation task. In Experiment 1, participants made judgments about the time interval between i) their button press and a rocket launch, and ii) a non-causal (...) predictive signal and rocket launch. In Experiment 2, an additional causal condition was included in which participants made judgments about the interval between an experimenter’s button press and the launch of a rocket. Temporal binding was demonstrated consistently and did not change in magnitude with age: estimates of delay were shorter in causal contexts for both adults and children. Additionally, the magnitude of the binding effect was greater when participants themselves were the cause of an outcome compared to when they were mere spectators. This suggests that although causality underlies the binding effect, intentional action may modulate its magnitude. Again, this was true of both adults and children. Taken together, these results are the first to suggest that the binding effect is present and developmentally constant from childhood into adulthood. (shrink)

Experiments have shown that prescriptive norms often influence causal inferences. The reason for this effect is still not clear. One problem of the studies is that the term ‘cause’ in the test questions is ambiguous and can refer to both the causal mechanism and the agent’s accountability. Possibly subjects interpreted the causal test question as a request to assess accountability rather than causality. Scenarios that put more stress on the causal mechanism should therefore yield no norm effect. Consequently, Experiment 1 (...) demonstrates that norms no longer influence causal judgments when the causal information is presented in a trial-by-trial learning task. Furthermore, Experiment 2 shows that norm effects are only obtained when the test question asks about a (potentially accountable) person but not when asked about a component of the causal mechanism. Both findings demonstrate that norms cease to influence causal judgments when the task settings highlight causal relations. (shrink)

Among the many causes of an event, how do we distinguish the important ones? Are there ways to distinguish among causes on principled grounds that integrate both practical aims and objective knowledge? Psychologist Tania Lombrozo has suggested that causal explanations “identify factors that are ‘exportable’ in the sense that they are likely to subserve future prediction and intervention” (Lombrozo 2010, 327). Hence portable causes are more important precisely because they provide objective information to prediction and intervention as practical aims. However, (...) I argue that this is only part of the epistemology of causal selection. Recent work on portable causes has implicitly assumed them to be portable within the same causal system at a later time. As a result, it has appeared that the objective content of causal selection includes only facts about the causal structure of that single system. In contrast, I present a case study from systems biology in which scientists are searching for causal factors that are portable across rather than within causal systems. By paying careful attention to how these biologists find portable causes, I show that the objective content of causal selection can extend beyond the immediate systems of interest. In particular, knowledge of the evolutionary history of gene networks is necessary for correctly identifying causal patterns in these networks that explain cellular behavior in a portable way. (shrink)

There is ample evidence that violations of injunctive norms impact ordinary causal attributions. This has struck some as deeply surprising, taking the ordinary concept of causation to be purely descriptive. Our explanation of the findings—the responsibility view—rejects this: we contend that the concept is in fact partly normative, being akin to concepts like responsibility and accountability. Based on this account, we predicted a very different pattern of results for causal attributions when an agent violates a statistical norm. And this pattern (...) has been borne out by the data. These predictions were based on the responsibility attributions that we would make. In this paper, I extend these previous findings, testing responsibility attributions. The results confirm the basis of our predictions, showing the same pattern of effects previously found for causal attributions for both injunctive norms and statistical norms. In fact, the results for responsibility attributions are not statistically significantly different from those previously found for causal attributions. I argue that this close correspondence lends further credence to the responsibility view over competing explanations of the impact of norms on causal attributions. (shrink)

Mary Shepherd (1777–1847) was a fierce and brilliant critic of Berkeley and Hume, who moreover offered strikingly original positive views about the nature of reality and our access to it which deserve much more attention (and credit, since she anticipates many prominent views) than they have received thus far. By way of illustration, I focus on Shepherd's 1824 Essay Upon the Relation of Cause and Effect, Controverting the Doctrine of Mr. Hume, Concerning the Nature of that Relation (ERCE). After a (...) brief setup, I canvas certain of her trenchant objections to Hume’s argumentation; I then present the positive core of her response to Hume, which consists in providing novel accounts of how reason alone or reason coupled with experience can justify, first, that every effect must have a cause, and second, that it is necessary that like causes produce like effects. Among other contributions here, Shepherd provides a distinctively metaphysical argument for the claim that nothing can begin to exist 'of itself' (going beyond an appeal to the Principle of Sufficient Reason, in particular), and leverages difference-making considerations to make the case that a single 'experimentum crucis' can justify causal belief (anticipating Mill's 'method of difference'). I close by highlighting salient features of Shepherd's metaphysics of causation, whereby causation is singularist and local (anticipating Ducasse and Anscombe) and involves synchronic interactions (anticipating Mill's and certain contemporary accounts), and according to which objects are essentially characterized by their causes and effects (anticipating contemporary causal or dispositional essentialist positions). (shrink)

This paper has two main parts. In the first part, we motivate a kind of indeterminate, suppositional credences by discussing the prospect for a subjective interpretation of a causal Bayesian network, an important tool for causal reasoning in artificial intelligence. A CBN consists of a causal graph and a collection of interventional probabilities. The subjective interpretation in question would take the causal graph in a CBN to represent the causal structure that is believed by an agent, and interventional probabilities in (...) a CBN to represent suppositional credences. We review a difficulty noted in the literature with such an interpretation, and suggest that a natural way to address the challenge is to go for a generalization of CBN that allows indeterminate credences. In the second part, we develop a decision-theoretic foundation for such indeterminate suppositional credences, by generalizing a theory of coherent choice functions to accommodate some form of act-state dependence. The upshot is a decision-theoretic framework that is not only rich enough to, so to speak, ground the probabilities in a subjectively interpreted causal network, but also interesting in its own right, in that it accommodates both act-state dependence and imprecise probabilities. (shrink)

This chapter presents a typology of the different kinds of inductive inferences we can draw from our evidence, based on the explanatory relationship between evidence and conclusion. Drawing on the literature on graphical models of explanation, I divide inductive inferences into (a) downwards inferences, which proceed from cause to effect, (b) upwards inferences, which proceed from effect to cause, and (c) sideways inferences, which proceed first from effect to cause and then from that cause to an additional effect. I further (...) distinguish between direct and indirect forms of downwards and upwards inferences. I then show how we can subsume canonical forms of inductive inference mentioned in the literature, such as inference to the best explanation, enumerative induction, and analogical inference, under this typology. Along the way, I explore connections with probability and confirmation, epistemic defeat, the relation between abduction and enumerative induction, the compatibility of IBE and Bayesianism, and theories of epistemic justification. (shrink)

"Correlation is not causation" is one of the mantras of the sciences—a cautionary warning especially to fields like epidemiology and pharmacology where the seduction of compelling correlations naturally leads to causal hypotheses. The standard view from the epistemology of causation is that to tell whether one correlated variable is causing the other, one needs to intervene on the system—the best sort of intervention being a trial that is both randomized and controlled. In this paper, we argue that some purely correlational (...) data contains information that allows us to draw causal inferences: statistical noise. Methods for extracting causal knowledge from noise provide us with an alternative to randomized controlled trials that allows us to reach causal conclusions from purely correlational data. (shrink)

People more frequently select norm-violating factors, relative to norm- conforming ones, as the cause of some outcome. Until recently, this abnormal-selection effect has been studied using retrospective vignette-based paradigms. We use a novel set of video stimuli to investigate this effect for prospective causal judgments—i.e., judgments about the cause of some future outcome. Four experiments show that people more frequently select norm- violating factors, relative to norm-conforming ones, as the cause of some future outcome. We show that the abnormal-selection effects (...) are not primarily explained by the perception of agency (Experiment 4). We discuss these results in relation to recent efforts to model causal judgment. (shrink)

When the absence of an event causes some outcome, it is an instance of omissive causation. For instance, not eating lunch may cause you to be hungry. Recent psychological proposals concur that the mind represents causal relations, including omissive causal relations, through mental simulation, but they disagree on the form of that simulation. One theory states that people represent omissive causes as force vectors; another states that omissions are representations of contrasting counterfactual simulations; a third argues that people think about (...) omissions by representing sets of iconic possibilities – mental models – in a piecemeal fashion. In this paper, we tease apart the empirical predictions of the three theories and describe experiments that run counter to two of them. Experiments 1 and 2 show that reasoners can infer temporal relations from omissive causes – a pattern that contravenes the force theory. Experiment 3 asked participants to list the possibilities consistent with an omissive cause – it found that they tended to list particular privileged possibilities first, most often, and faster than alternative possibilities. The pattern is consistent with the model theory, but inconsistent with the contrast hypothesis. We marshal the evidence and explain why it helps to solve a long-standing debate about how the mind represents omissions. (shrink)

Current debates surrounding the virtues and shortcomings of randomization are symptomatic of a lack of appreciation of the fact that causation can be inferred by two distinct inference methods, each requiring its own, specific experimental design. There is a non-statistical type of inference associated with controlled experiments in basic biomedical research; and a statistical variety associated with randomized controlled trials in clinical research. I argue that the main difference between the two hinges on the satisfaction of the comparability requirement, which (...) is in turn dictated by the nature of the objects of study, namely homogeneous or heterogeneous populations of biological systems. Among other things, this entails that the objection according to which randomized experiments fail to provide better evidence for causation because randomization cannot guarantee comparability is mistaken. (shrink)

In this paper, I examine Cicero’s oft-neglected De Divinatione, a dialogue investigating the legitimacy of the practice of divination. First, I offer a novel analysis of the main arguments for divination given by Quintus, highlighting the fact that he employs two logically distinct argument forms. Next, I turn to the first of the main arguments against divination given by Marcus. Here I show, with the help of modern probabilistic tools, that Marcus’ skeptical response is far from the decisive, proto-naturalistic assault (...) on superstition that it is sometimes portrayed to be. Then, I offer an extended analysis of the second of the main arguments against divination given by Marcus. Inspired by Marcus’ second main argument, I formulate, explicate, and defend a substantive principle of scientific methodology that I call the “Ciceronian Causal-Nomological Requirement” (CCR). Roughly, this principle states that causal knowledge is essential for relying on correlations in predictive inference. Although I go on to argue that Marcus’ application of the CCR in his debate with Quintus is dialectically inadequate, I conclude that De Divinatione deserves its place in Cicero’s philosophical corpus, and that ultimately, its significance for the history and philosophy of science ought to be recognized. (shrink)

There has been increasing interest in recent years in the development of approaches to estimate causal effects when the number of potential confounders is prohibitively large. This growth in interest has led to a number of potential estimators one could use in this setting. Each of these estimators has different operating characteristics, and it is unlikely that one estimator will outperform all others across all possible scenarios. Coupling this with the fact that an analyst can never know which approach is (...) best for their particular data, we propose a synthetic estimator that averages over a set of candidate estimators. Averaging is widely used in statistics for problems such as prediction, where there are many possible models, and averaging can improve performance and increase robustness to using incorrect models. We show that these ideas carry over into the estimation of causal effects in high-dimensional scenarios. We show theoretically that averaging provides robustness against choosing a bad model, and show empirically via simulation that the averaging estimator performs quite well, and in most cases nearly as well as the best among all possible candidate estimators. Finally, we illustrate these ideas in an environmental wide association study and see that averaging provides the largest benefit in the more difficult scenarios that have large numbers of confounders. (shrink)

During the first half of the twentieth century, many philosophers of memory opposed the postulation of memory traces based on the claim that a satisfactory account of remembering need not include references to causal processes involved in recollection. However, in 1966, an influential paper by Martin and Deutscher showed that causal claims are indeed necessary for a proper account of remembering. This, however, did not settle the issue, as in 1977 Malcolm argued that even if one were to buy Martin (...) and Deutscher’s argument for causal claims, we still don’t need to postulate the existence of memory traces. This paper reconstructs the dialectic between realists and anti-realists about memory traces, suggesting that ultimately realists’ arguments amount to inferences to the best explanation. I then argue that Malcolm’s anti-realist strategy consists in the suggestion that causal explanations that do not invoke memory traces are at least as good as those that do. But then, Malcolm, I argue that there are a large number of memory phenomena for which explanations that do not postulate the existence of memory traces are definitively worse than explanations that do postulate them. Next, I offer a causal model based on an interventionist framework to illustrate when memory traces can help to explain memory phenomena and proceed to substantiate the model with details coming from extant findings in the neuroscience of memory. (shrink)

Finding causes is a central goal in psychological research. In this paper, I argue based on the interventionist approach to causal discovery that the search for psychological causes faces great obstacles. Psychological interventions are likely to be fat-handed: they change several variables simultaneously, and it is not known to what extent such interventions give leverage for causal inference. Moreover, due to problems of measurement, the degree to which an intervention was fat-handed, or more generally, what the intervention in fact did, (...) is difficult to reliably estimate. A further complication is that the causal findings in psychology are typically made at the population level, and such findings do not allow inferences to individual-level causal relationships. I also discuss the implications of these problems for research, as well as various ways of addressing them, such as focusing more on the discovery of robust but non-causal patterns. (shrink)

Within the field of causal inference, it is desirable to learn the structure of causal relationships holding between a system of variables from the correlations that these variables exhibit; a sub-problem of which is to certify whether or not a given causal hypothesis is compatible with the observed correlations. A particularly challenging setting for assessing causal compatibility is in the presence of partial information; i.e. when some of the variables are hidden/latent. This paper introduces the possible worlds framework as a (...) method for deciding causal compatibility in this difficult setting. We define a graphical object called a possible worlds diagram, which compactly depicts the set of all possible observations. From this construction, we demonstrate explicitly, using several examples, how to prove causal incompatibility. In fact, we use these constructions to prove causal incompatibility where no other techniques have been able to. Moreover, we prove that the possible worlds framework can be adapted to provide a complete solution to the possibilistic causal compatibility problem. Even more, we also discuss how to exploit graphical symmetries and cross-world consistency constraints in order to implement a hierarchy of necessary compatibility tests that we prove converges to sufficiency. (shrink)

Many animal species use tools, but human technical engagement is more complex. We argue that there is coevolution between technical engagement and advanced forms of causal cognition in the human lineage. As an analytic tool, we present a classification of different forms of causal thinking. Human causal thinking has become detached from space and time, so that instead of just reacting to perceptual input, our minds can simulate actions and forces and their causal consequences. Our main thesis is that, unlike (...) the situation for other primate species, an increasing emphasis on technical engagement made some hominins capable of reasoning about the forces involved in causal processes. This thesis is supported in three ways: We compare the casual thinking about forces of hominins with that of other primates. We analyze the causal thinking required for Stone Age hunting technologies such as throwing spears, bow hunting and the use of poisoned arrows, arguing that they may serve as examples of the expansion of casual cognition about forces. We present neurophysiological results that indicate the facilitation of advanced causal thinking. (shrink)

As the demand for explainable deep learning grows in the evaluation of language technologies, the value of a principled grounding for those explanations grows as well. Here we study the state-of-the-art in explanation for neural models for natural-language processing (NLP) tasks from the viewpoint of philosophy of science. We focus on recent evaluation work that finds brittleness in explanations obtained through attention mechanisms.We harness philosophical accounts of explanation to suggest broader conclusions from these studies. From this analysis, we assert the (...) impossibility of causal explanations from attention layers over text data. We then introduce NLP researchers to contemporary philosophy of science theories that allow robust yet non-causal reasoning in explanation, giving computer scientists a vocabulary for future research. (shrink)

In temporal binding, the temporal interval between one event and another, occurring some time later, is subjectively compressed. We discuss two ways in which temporal binding has been conceptualized. In studies showing temporal binding between a voluntary action and its causal consequences, such binding is typically interpreted as providing a measure of an implicit or pre-reflective “sense of agency”. However, temporal binding has also been observed in contexts not involving voluntary action, but only the passive observation of a cause-effect sequence. (...) In those contexts, it has been interpreted as a top-down effect on perception reflecting a belief in causality. These two views need not be in conflict with one another, if one thinks of them as concerning two separate mechanisms through which temporal binding can occur. In this paper, we explore an alternative possibility: that there is a unitary way of explaining temporal binding both within and outside the context of voluntary action as a top-down effect on perception reflecting a belief in causality. Any such explanation needs to account for ways in which agency, and factors connected with agency, have been shown to affect the strength of temporal binding. We show that principles of causal inference and causal selection already familiar from the literature on causal learning have the potential to explain why the strength of people’s causal beliefs can be affected by the extent to which they are themselves actively involved in bringing about events, thus in turn affecting binding. (shrink)

While the HVTN 505 trial showed no overall efficacy of the tested vaccine to prevent HIV infection over placebo, markers measuring immune response to vaccination were strongly correlated with infection. This finding generated the hypothesis that some marker-defined vaccinated subgroups were partially protected whereas others had their risk increased. This hypothesis can be assessed using the principal stratification framework for studying treatment effect modification by an intermediate response variable, using methods in the sub-field of principal surrogate analysis that studies multiple (...) principal strata. Unfortunately, available methods for PS analysis require an augmented study design not available in HVTN 505, and make untestable structural risk assumptions, motivating a need for more robust PS methods. Fortunately, another sub-field of principal stratification, survivor average causal effect analysis – which studies effects in a single principal stratum – provides many methods not requiring an augmented design and making fewer assumptions. We show how, for a binary intermediate response variable, methods developed for SACE analysis can be adapted to PS analysis, providing new and more robust PS methods. Application to HVTN 505 supports that the vaccine partially protected individuals with vaccine-induced T-cells expressing certain combinations of functions. (shrink)

Discovering high-level causal relations from low-level data is an important and challenging problem that comes up frequently in the natural and social sciences. In a series of papers, Chalupka etal. (2015, 2016a, 2016b, 2017) develop a procedure forcausal feature learning (CFL) in an effortto automate this task. We argue that CFL does not recommend coarsening in cases where pragmatic considerations rule in favor of it, and recommends coarsening in cases where pragmatic considerations rule against it. We propose a new technique, (...) pragmatic causal feature learning (PCFL), which extends the original CFL algorithm in useful and intuitive ways. We show that PCFL has the same attractive measure-theoretic properties as the original CFL algorithm. We compare the performance of both methods through theoretical analysis and experiments. (shrink)

Within Non-monotonic Reasoning, numerous principles of causal reasoning have been proposed. Many of these principles have been viewed as desirable in formalisms that reason with causality, and have been widely adopted throughout the literature. We provide a critique of these principles, evaluate their suitability for characterising and formulating causal non-monotonic reasoning, and find that most are unsuitable. Further, we discuss a new approach to causal non-monotonic reasoning motivated by how humans typically reason with causality.

An intervention may have an effect on units other than those to which it was administered. This phenomenon is called interference and it usually goes unmodeled. In this paper, we propose to combine Lauritzen-Wermuth-Frydenberg and Andersson-Madigan-Perlman chain graphs to create a new class of causal models that can represent both interference and non-interference relationships for Gaussian distributions. Specifically, we define the new class of models, introduce global and local and pairwise Markov properties for them, and prove their equivalence. We also (...) propose an algorithm for maximum likelihood parameter estimation for the new models, and report experimental results. Finally, we show how to compute the effects of interventions in the new models. (shrink)

Quantum mechanics admits a “linguistic interpretation” if one equates preliminary any quantum state of some whether quantum entity or word, i.e. a wave function interpret-able as an element of the separable complex Hilbert space. All possible Feynman pathways can link to each other any two semantic units such as words or term in any theory. Then, the causal reasoning would correspond to the case of classical mechanics (a single trajectory, in which any next point is causally conditioned), and the probabilistic (...) reasoning, to the case of quantum mechanics (many Feynman trajectories). Frame semantics turns out to be the natural counterpart of that linguistic interpretation of quantum mechanics. (shrink)

The COVID-19 pandemic has demonstrated how different (even conflicting) interventions on nature can be scientifically justified: interventions can be deemed "effective" only in relation to specific target variables - in relation to variables the values of which we seek to control. Choosing the "right" target variables, in turn, depends on our values and pragmatic aims. This essay is based on a presentation given at the symposium "Multidisciplinary Perspectives on the COVID-19 Pandemic", organised at the Helsinki Collegium for Advanced Studies on (...) 16 June 2020. (shrink)

This article provides a conceptual account of causal understanding by connecting current psychological research on time and causality with philosophical debates on the causal asymmetry. I argue that causal relations are viewed as asymmetric because they are understood in temporal terms. I investigate evidence from causal learning and reasoning in both children and adults: causal perception, the temporal priority principle, and the use of temporal cues for causal inference. While this account does not suffice for correct inferences of causal structure, (...) I show it to serve as a preliminary understanding of causal concepts as asymmetric, that later incorporates other types of evidence. This approach supplies causal models with an asymmetric concept of causation that underlies hypotheses about causal structure, as I will illustrate from the framework of the knowledge-based causal induction model. I further argue for an integrating perspective, showing how the understanding of causes as preceding their effects underlies both psychological models and philosophical debates over time and the causal asymmetry, particularly regarding problem cases such as simultaneous causation or backwards causation, and the conceptual connection between causation and action. (shrink)

Randomized control trials are the gold standard for estimating causal effects, but often use samples that are non-representative of the actual population of interest. We propose a reweighting method for estimating population average treatment effects in settings with noncompliance. Simulations show the proposed compliance-adjusted population estimator outperforms its unadjusted counterpart when compliance is relatively low and can be predicted by observed covariates. We apply the method to evaluate the effect of Medicaid coverage on health care use for a target population (...) of adults who may benefit from expansions to the Medicaid program. We draw RCT data from the Oregon Health Insurance Experiment, where less than one-third of those randomly selected to receive Medicaid benefits actually enrolled. (shrink)

I look at the two main approaches used to count COVID-19 deaths and show how each of those approaches can appear to both overcount COVID deaths (including deaths it should exclude) and undercount COVID deaths (excluding deaths it should include). I trace this to the fact - well-known to philosophers - that causal attribution is interest-relative. Which deaths we should attribute to COVID (as opposed to other causes) will depend on our particular interests and values. Contrary to what many journalists (...) and researchers report, there is therefore no such thing as the "true" COVID death toll. Understanding this can help us to become more sophisticated consumers of COVID information. I conclude by suggesting that scientists, by reflecting on society's varied interests and values, are in a position to construct measures of COVID-caused mortality that are potentially as useful, or more useful, than the measures we currently have. (shrink)

Causal queries about singular cases, which inquire whether specific events were causally connected, are prevalent in daily life and important in professional disciplines such as the law, medicine, or engineering. Because causal links cannot be directly observed, singular causation judgments require an assessment of whether a co‐occurrence of two events c and e was causal or simply coincidental. How can this decision be made? Building on previous work by Cheng and Novick (2005) and Stephan and Waldmann (2018), we propose a (...) computational model that combines information about the causal strengths of the potential causes with information about their temporal relations to derive answers to singular causation queries. The relative causal strengths of the potential cause factors are relevant because weak causes are more likely to fail to generate effects than strong causes. But even a strong cause factor does not necessarily need to be causal in a singular case because it could have been preempted by an alternative cause. We here show how information about causal strength and about two different temporal parameters, the potential causes' onset times and their causal latencies, can be formalized and integrated into a computational account of singular causation. Four experiments are presented in which we tested the validity of the model. The results showed that people integrate the different types of information as predicted by the new model. (shrink)

Although it has long been known that time is a cue to causation, recent work with adults has demonstrated that causality can also influence the experience of time. In causal reordering (Bechlivanidis & Lagnado, 2013, 2016) adults tend to report the causally consistent order of events, rather than the correct temporal order. However, the effect has yet to be demonstrated in children. Across four pre-registered experiments, 4- to 10-year-old children (N=813) and adults (N=178) watched a 3-object Michotte-style ‘pseudocollision’. While in (...) the canonical version of the clip object A collided with B, which then collided with object C (order: ABC), the pseudocollision involved the same spatial array of objects but featured object C moving before object B (order: ACB), with no collision between B and C. Participants were asked to judge the temporal order of events and whether object B collided with C. Across all age groups, participants were significantly more likely to judge that B collided with C in the 3-object pseudocollision than in a 2-object control clip (where clear causal direction was lacking), despite the spatiotemporal relations between B and C being identical in the two clips (Experiments 1—3). Collision judgements and temporal order judgements were not entirely consistent, with some participants—particularly in the younger age range—basing their temporal order judgements on spatial rather than temporal information (Experiment 4). We conclude that in both children and adults, rather than causal impressions being determined only by the basic spatial-temporal properties of object movement, schemata are used in a top-down manner when interpreting perceptual displays. (shrink)

The causal compatibility question asks whether a given causal structure graph — possibly involving latent variables — constitutes a genuinely plausible causal explanation for a given probability distribution over the graph’s observed categorical variables. Algorithms predicated on merely necessary constraints for causal compatibility typically suffer from false negatives, i.e. they admit incompatible distributions as apparently compatible with the given graph. In 10.1515/jci-2017-0020, one of us introduced the inflation technique for formulating useful relaxations of the causal compatibility problem in terms of (...) linear programming. In this work, we develop a formal hierarchy of such causal compatibility relaxations. We prove that inflation is asymptotically tight, i.e., that the hierarchy converges to a zero-error test for causal compatibility. In this sense, the inflation technique fulfills a longstanding desideratum in the field of causal inference. We quantify the rate of convergence by showing that any distribution which passes the nth-order inflation test must be On−1/2$\begin{array}{} \displaystyle {O}{\left} \end{array}$-close in Euclidean norm to some distribution genuinely compatible with the given causal structure. Furthermore, we show that for many causal structures, the causal compatibility problem is faithfully formulated already by either the first or second order inflation test. (shrink)

McCaffrey and Danks have posed the challenge of discovering causal relations in data drawn from a mixture of distributions as an impossibility result in functional magnetic resonance. We give an algorithm that addresses this problem for the distributions commonly assumed in fMRI studies and find that in testing, it can accurately separate data from mixed distributions. As with other obstacles to automated search, the problem of mixed distributions is not an impossible one, but rather a challenge. 1Introduction2Background3Addressing the Problem4Discussion.