Inferring causal networks from observations and interventions

Information about the structure of a causal system can come in the form of observational data—random samples of the system's autonomous behavior—or interventional data—samples conditioned on the particular values of one or more variables that have been experimentally manipulated. Here we study people's ability to infer causal structure from both observation and intervention, and to choose informative interventions on the basis of observational data. In three causal inference tasks, participants were to some degree capable of distinguishing between competing causal hypotheses on the basis of purely observational data. Performance improved substantially when participants were allowed to observe the effects of interventions that they performed on the systems. We develop computational models of how people infer causal structure from data and how they plan intervention experiments, based on the representational framework of causal graphical models and the inferential principles of optimal Bayesian decision‐making and maximizing expected information gain. These analyses suggest that people can make rational causal inferences, subject to psychologically reasonable representational assumptions and computationally reasonable processing constraints.     

Reasons

The temptation to look for the “purely normative essence” of argument stems from the understandable ambition to distinguish rational persuasion from mere persuasion. But in seeking a purely normative notion of argument it is easy to overlook—or actually deny—that rational persuasion is a kind of persuasion. The burden of this essay is to show that the concept of reason from which our interest in argument derives can only exist and have normative force as a kind of persuasion, that is, as something (also) causal.

     

Two Types of Quidditism

According to structuralism, all natural properties are individuated by their roles in causal/nomological structures. According to quidditism, at least some natural properties are individuated in some other way. Because these theses deal with the identities of natural properties, this distinction cuts to the core of a serious metaphysical dispute: Are the intrinsic natures of all natural properties essentially causal/nomological in character? I'll argue that the answer is ‘no’, or at least that this answer is more plausible than many critics of quidditism have recognized. In section 1, I distinguish between two versions of quidditism. Bare quidditism holds that worlds with distinct properties and isomorphic structures must be qualitatively identical in the following sense: inhabiting one world would be indistinguishable from inhabiting the other. In contrast, qualitative quidditism allows such worlds to have qualitative differences. In section 2, I discuss an epistemological position that allows us to better understand the sense in which isomorphic structures can be qualitatively distinct. In sections 3 and 4, I argue that qualitative quidditism fares much better than bare quidditism with respect to a number of well-known arguments in the field—both for quidditism and against it.     

Correlational Data,Causal Hypotheses,and Validity

A shared problem across the sciences is to make sense of correlational data coming from observations and/or from experiments. Arguably, this means establishing when correlations are causal and when they are not. This is an old problem in philosophy. This paper, narrowing down the scope to quantitative causal analysis in social science, reformulates the problem in terms of the validity of statistical models. Two strategies to make sense of correlational data are presented: first, a ‘structural strategy’, the goal of which is to model and test causal structures that explain correlational data; second, a ‘manipulationist or interventionist strategy’, that hinges upon the notion of invariance under intervention. It is argued that while the former can offer a solution the latter cannot.     

What causal conditional reasoning tells us about people's understanding of causality

Causal conditional reasoning means reasoning from a conditional statement that refers to causal content. We argue that data from causal conditional reasoning tasks tell us something not only about how people interpret conditionals, but also about how they interpret causal relations. In particular, three basic principles of people's causal understanding emerge from previous studies: the modal principle, the exhaustive principle, and the equivalence principle. Restricted to the four classic conditional inferences—Modus Ponens, Modus Tollens, Denial of the Antecedent, and Affirmation of the Consequent—causal conditional reasoning data are only partially able to support these principles. We present three experiments that use concrete and abstract causal scenarios and combine inference tasks with a new type of task in which people reformulate a given causal situation. The results provide evidence for the proposed representational principles. Implications for theories of the naïve understanding of causality are discussed.     

Applying cognitive training to target executive functions during early development

Developmental psychopathology is increasingly recognizing the importance of distinguishing causal processes (i.e., the mechanisms that cause a disease) from developmental outcomes (i.e., the symptoms of the disorder as it is eventually diagnosed). Targeting causal processes early in disordered development may be more effective than waiting until outcomes are established and then trying to reverse the pathogenic process. In this review, I evaluate evidence suggesting that neural and behavioral plasticity may be greatest at very early stages of development. I also describe correlational evidence suggesting that, across a number of conditions, early emerging individual differences in attentional control and working memory may play a role in mediating later-developing differences in academic and other forms of learning. I review the currently small number of studies that applied direct and indirect cognitive training targeted at young individuals and discuss methodological challenges associated with targeting this age group. I also discuss a number of ways in which early, targeted cognitive training may be used to help us understand the developmental mechanisms subserving typical and atypical cognitive development.     

Causal Networks or Causal Islands? The Representation of Mechanisms and the Transitivity of Causal Judgment

Knowledge of mechanisms is critical for causal reasoning. We contrasted two possible organizations of causal knowledge—an interconnected causal network, where events are causally connected without any boundaries delineating discrete mechanisms; or a set of disparate mechanisms—causal islands—such that events in different mechanisms are not thought to be related even when they belong to the same causal chain. To distinguish these possibilities, we tested whether people make transitive judgments about causal chains by inferring, given A causes B and B causes C, that A causes C. Specifically, causal chains schematized as one chunk or mechanism in semantic memory (e.g., exercising, becoming thirsty, drinking water) led to transitive causal judgments. On the other hand, chains schematized as multiple chunks (e.g., having sex, becoming pregnant, becoming nauseous) led to intransitive judgments despite strong intermediate links ((Experiments 1–3). Normative accounts of causal intransitivity could not explain these intransitive judgments (Experiments 4 and 5).     

Causal judgments about relations between multilevel variables

Three experiments presented stimulus information about cause and effect variables taking 3 quantitative values. Judgments tended to vary in accordance with considerations of conditions affecting the validity of causal inference from correlational data: whether causal candidates were presented simultaneously or in a temporal order such that one could affect the other and whether candidates were confounded with each other. The results supported a general hypothesis that causal judgments are moderated in accordance with acquired methodological intuitions. The fourth experiment showed that tendencies in correlation judgment were different from those in causal judgment, further supporting the hypothesis that causal judgment from multilevel variable information is, to some extent, determined by processes or conceptual frameworks specific to the domain of causal cognition.     

Multiplicative invalidity and its application to complex correlational models

An ever-increasing proportion of social psychology researchers use various versions of complex correlational models such as path analyses or structural equation models and others to draw causal conclusions from correlational data. Critics of complex correlational models have pointed out that (a) misspecification errors are the rule rather than the exception, (b) one cannot draw causal conclusions from a set of correlations, (c) most researchers fail to adjust their correlations for attenuation due to unreliability, and (d) the measures researchers use may actually be measures of outside variables that are correlated with other variables in one's model. Rather than rehash the debates that go along with these criticisms, the author makes some assumptions that are extremely favorable to the complex correlational modeler in that all of these criticisms are disallowed. Nevertheless, even with these assumptions, the author shows how spurious direct and indirect effects are likely to be created by moderately valid measures when researchers compute complex correlations. The author concludes that until social psychologists are better able to deal with the issue of the validity of their measures, they should not use complex correlational models.     

Self‐Stigma of Mental Illness and Help Seeking Among School Counselors

The authors reported the findings from a correlational investigation examining the relationship between school counselors' (N = 333) self‐stigma of mental illness, help‐seeking behaviors, burnout, stress, and life satisfaction. The authors used a path analysis to test a hypothesized causal framework that self‐stigma of mental illness contributed to help‐seeking behaviors, which contributed to stress and burnout. The findings showed that the data fit the hypothesized model. The authors discuss the implications of the findings.     

Controlled Substance‐related Beliefs and Use: Relationships to Undergraduates' Creative Personality Traits

The relationship between creativity and various mind‐altering substances — especially alcohol — has been a popular topic among creativity researchers and the public at large. Yet experimental studies have found little evidence that alcohol use has a causal influence on creativity, with most studies of creative production showing negative or neutral effects. However, the impact of high levels of creativity on future alcohol use (i.e., the reciprocal relationship) has been rarely studied. The present study examined the relationship between creative personality characteristics, use of alcohol, tobacco, and marijuana, and alcohol‐related beliefs among 431 undergraduates. Results of correlational and t‐test analyses suggest that creative personality is not significantly correlated with use of these three controlled substances, and that the relationships between personality scores and self‐reported beliefs about alcohol are generally weak, with notable exceptions. The paper concludes with specific recommendations regarding research on the relationship between creativity and the use of controlled‐substances.     

Feature Statistics Modulate the Activation of Meaning During Spoken Word Processing

Understanding spoken words involves a rapid mapping from speech to conceptual representations. One distributed feature‐based conceptual account assumes that the statistical characteristics of concepts’ features—the number of concepts they occur in (distinctiveness/sharedness) and likelihood of co‐occurrence (correlational strength)—determine conceptual activation. To test these claims, we investigated the role of distinctiveness/sharedness and correlational strength in speech‐to‐meaning mapping, using a lexical decision task and computational simulations. Responses were faster for concepts with higher sharedness, suggesting that shared features are facilitatory in tasks like lexical decision that require access to them. Correlational strength facilitated responses for slower participants, suggesting a time‐sensitive co‐occurrence‐driven settling mechanism. The computational simulation showed similar effects, with early effects of shared features and later effects of correlational strength. These results support a general‐to‐specific account of conceptual processing, whereby early activation of shared features is followed by the gradual emergence of a specific target representation.     

Developmental psychology in the real world: A paradigm of parent education

Twenty years ago, William Kessen, Greta Fein and I developed and tested a model of parent education, a model which involved variation in curricular content, didactic approaches, and child outcomes–experimentally contrasted. Our experience yielded some valuable lessons about the extent to which researchers can influence parents' behaviour, and parents their children's development. In the present article I suggest that these lessons might be useful for researchers now as they were for us then. I present some observations about the research that developmental psychologists have done in the intervening two decades and suggest that we could learn much by using such a model of parent education to answer questions about the effets of parents' behaviour on children's development. Carefully done, parent education studies can yield valuable information about many of the questions raised in recent correlational research. Parent training research can suggest hypotheses that can be tested with data collected non-interventively and analysed in causal model analyses–and vice versa. What is more, because one reason for studying parent effects is so that we can offer guidance to parents about rearing their children, this design has a particular advantage; it is both the medium and the message.     

Hearing Silent Shapes: Identifying the Shape of a Sound-Obstructing Surface

Two experiments test whether the shape of objects that obstruct sound can be perceived by human listeners. Three foam-core shapes of equal area—disk, square, and triangle—were positioned in front a set of loudspeakers, which emanated broadband noise. On each trial, blindfolded listeners were asked to identify which shape obstructed the noise. Both experiments revealed that under most conditions, listeners could identify the shapes at better-than-chance levels. Experiment 2 also showed that the addition of a second intensity level of broadband noise randomized across trials actually improved performance. This finding suggests that listeners were likely basing their judgments on an acoustic dimension that was invariant—and was perhaps made more salient—over multiple intensities. These results add to the growing literature showing that human listeners are sensitive to sound-structuring surfaces that themselves do not produce sound.     

Neural Synchrony and the Causal Efficacy of Consciousness

The purpose of this paper is to address a well-known dilemma for physicalism. If mental properties are type identical to physical properties, then their causal efficacy is secure, but at the cost of ruling out mentality in creatures very different to ourselves. On the other hand, if mental properties are multiply realizable, then all kinds of creatures can instantiate them, but then they seem to be causally redundant. The causal exclusion problem depends on the widely held principle that realized properties inherit their causal powers from their realizers. While this principle holds for functional realization, it fails on a broader notion of realization that permits the realization of complex qualitative properties such as spatial and temporal patterns. Such properties are best seen as dependent powerful qualities, which have their causal roles in virtue of being the qualities they are, and do not inherit powers from their realizers. Recent studies have identified one such property—neural synchrony—as a correlate of consciousness. If synchrony is also partially constitutive of consciousness, then phenomenal properties are both multiply realizable and causally novel. I outline a version of representationalism about consciousness on which this constitution claim holds.

     

Book Reviews

Book reviewed in this article: Contrasts and effect sizes in behavioral research. A correlational approach: By R. Rosenthal, R. L. Rosnow, & D. R. Rubin Multilevel modeling: Methodological advances, issues and applications: Edited by S. P. Reise & N. Duan The mind's arrows—Bayes nets and graphical causal models in psychology: By C. Glymour Psychological research: The ideas behind the methods: By Douglas G. Mook Ordinal measurement in the behavioral sciences: By Norman Cliff and John A. Keats Introduction to nonparametric item response theory: By K. Sijtsma and I. W. Molenaar The analysis and interpretation of multivariate data for social scientists: By D. J. Bartholomew, F. Steele, I. Moustaki, & J. I. Galbraith     

MICRO-Foundations in Strategic Management: Squaring Coleman’s Diagram

Abell, Felin and Foss argue that “macro-explanations” in strategic management, explanations in which organizational routines figure prominently and in which both the explanandum and explanans are at the macro-level, are necessarily incomplete. They take a diagram (which has the form of a trapezoid) from Coleman, Foundations of Social Theory, The Belknap Press of Harvard University Press, Cambridge (Mass.)/London, (1990) to task to show that causal chains connecting two macro-phenomena always involve “macro-to-micro” and “micro-to-macro” links, links that macro-explanations allegedly fail to recognize. Their plea for micro-foundations in strategic management is meant to shed light on these “missing links”. The paper argues that while there are good reasons for providing micro-foundations, Abell, Felin and Foss’s causal incompleteness argument is not one of them. Their argument does not sufficiently distinguish between causal and constitutive relations. Once these relations are carefully distinguished, it follows that Coleman’s diagram has to be squared. This in turn allows us to see that macro-explanations need not be incomplete.     

Reflective Equilibrium as an Ameliorative Framework for Feminist Epistemology

As Helen Longino's overview of Hypatia's engagement with feminist epistemology suggests, the last twenty‐five years’ contributions to this field reveal a strong focus on the topic of knowledge. In her short outline, Longino questions this narrow focus on knowledge in epistemological inquiry. The main purpose of this article is to provide a framework for systematically taking up the questions raised by Longino, one that prevents us from running the risk of becoming unreflectively involved in sexist, racist, or otherwise problematic inquiry. I argue that a specific form of the method of Reflective Equilibrium, as it is widely discussed in moral epistemology, logic, and theories of rationality, enables us to cope with the problems of traditional epistemology, which feminist theorizers such as Sally Haslanger have pointed to. With the account of Reflective Equilibrium I am offering—drawing in many respects on the model provided by Catherine Z. Elgin—we have an ameliorative method that allows us to rethink epistemological values, goals, and standards in a systematic way, and that largely avoids implicit and explicit biases in epistemology.     

Learning a theory of causality

The very early appearance of abstract knowledge is often taken as evidence for innateness. We explore the relative learning speeds of abstract and specific knowledge within a Bayesian framework and the role for innate structure. We focus on knowledge about causality, seen as a domain-general intuitive theory, and ask whether this knowledge can be learned from co-occurrence of events. We begin by phrasing the causal Bayes nets theory of causality and a range of alternatives in a logical language for relational theories. This allows us to explore simultaneous inductive learning of an abstract theory of causality and a causal model for each of several causal systems. We find that the correct theory of causality can be learned relatively quickly, often becoming available before specific causal theories have been learned--an effect we term the blessing of abstraction. We then explore the effect of providing a variety of auxiliary evidence and find that a collection of simple perceptual input analyzers can help to bootstrap abstract knowledge. Together, these results suggest that the most efficient route to causal knowledge may be to build in not an abstract notion of causality but a powerful inductive learning mechanism and a variety of perceptual supports. While these results are purely computational, they have implications for cognitive development, which we explore in the conclusion.