The role of causal models in analogical inference.

Computational models of analogy have assumed that the strength of an inductive inference about the target is based directly on similarity of the analogs and in particular on shared higher order relations. In contrast, work in philosophy of science suggests that analogical inference is also guided by causal models of the source and target. In 3 experiments, the authors explored the possibility that people may use causal models to assess the strength of analogical inferences. Experiments 1-2 showed that reducing analogical overlap by eliminating a shared causal relation (a preventive cause present in the source) from the target increased inductive strength even though it decreased similarity of the analogs. These findings were extended in Experiment 3 to cross-domain analogical inferences based on correspondences between higher order causal relations. Analogical inference appears to be mediated by building and then running a causal model. The implications of the present findings for theories of both analogy and causal inference are discussed. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Expectations and interpretations during causal learning.

In existing models of causal induction, 4 types of covariation information (i.e., presence/absence of an event followed by presence/absence of another event) always exert identical influences on causal strength judgments (e.g., joint presence of events always suggests a generative causal relationship). In contrast, we suggest that, due to expectations developed during causal learning, learners give varied interpretations to covariation information as it is encountered and that these interpretations influence the resulting causal beliefs. In Experiments 1A–1C, participants' interpretations of observations during a causal learning task were dynamic, expectation based, and, furthermore, strongly tied to subsequent causal judgments. Experiment 2 demonstrated that adding trials of joint absence or joint presence of events, whose roles have been traditionally interpreted as increasing causal strengths, could result in decreased overall causal judgments and that adding trials where one event occurs in the absence of another, whose roles have been traditionally interpreted as decreasing causal strengths, could result in increased overall causal judgments. We discuss implications for traditional models of causal learning and how a more top-down approach (e.g., Bayesian) would be more compatible with the current findings. (PsycINFO Database Record (c) 2011 APA, all rights reserved)     

Campbell and Rubin: A primer and comparison of their approaches to causal inference in field settings.

This article compares Donald Campbell’s and Donald Rubin’s work on causal inference in field settings on issues of epistemology, theories of cause and effect, methodology, statistics, generalization, and terminology. The two approaches are quite different but compatible, differing mostly in matters of bandwidth versus fidelity. Campbell’s work demonstrates broad narrative scope that covers a wide array of concepts related to causation, with a powerful appreciation for human fallibility in making causal judgments, with a more elaborate theory of cause and generalization, and with a preference for design over analysis. Rubin’s approach is a more narrow and formal quantitative analysis of effect estimation, sharing a preference for design but best known for analysis, with compelling quantitative approaches to obtaining unbiased quantitative effect estimates from nonrandomized designs and with comparatively little to say about generalization. Much could be gained by joining the emphasis on design in Campbell with the emphasis on analysis in Rubin. However, the 2 approaches also speak modestly different languages that leave some questions about their total commensurability that only continued dialogue can fully clarify. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

A symbolic-connectionist theory of relational inference and generalization.

The authors present a theory of how relational inference and generalization can be accomplished within a cognitive architecture that is psychologically and neurally realistic. Their proposal is a form of symbolic connectionism: a connectionist system based on distributed representations of concept meanings, using temporal synchrony to bind fillers and roles into relational structures. The authors present a specific instantiation of their theory in the form of a computer simulation model, Learning and Inference with Schemas and Analogies (LISA). By using a kind of self-supervised learning, LISA can make specific inferences and form new relational generalizations and can hence acquire new schemas by induction from examples. The authors demonstrate the sufficiency of the model by using it to simulate a body of empirical phenomena concerning analogical inference and relational generalization. (PsycINFO Database Record (c) 2011 APA, all rights reserved)     

Analogical problem solving: A hierarchical analysis of procedural similarity.

Similarity between source analogues and target problems is a central theme in the research on analogical transfer. Much of the theorizing and research has focused on the effects of superficial and structural similarity on transfer. The present research is an attempt to analyze systematically another critical type of similarity, namely, procedural similarity, and to examine its effects on the executing process. Participants viewed a schematic picture as a source model, interpreted its conceptual meaning, and then attempted to solve a problem to which the conceptual information from the source model could be applied. The results indicate that the ease with which a source solution was implemented was largely determined by the abstraction level at which a solution was shared by a source analogue and the target problem. The degree of procedural similarity was also found to influence the executing process in analogical transfer. A conceptual model concerning the function of procedural similarity as a utilizational constraint in analogical problem solving is proposed. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

A general approach to causal mediation analysis.

Traditionally in the social sciences, causal mediation analysis has been formulated, understood, and implemented within the framework of linear structural equation models. We argue and demonstrate that this is problematic for 3 reasons: the lack of a general definition of causal mediation effects independent of a particular statistical model, the inability to specify the key identification assumption, and the difficulty of extending the framework to nonlinear models. In this article, we propose an alternative approach that overcomes these limitations. Our approach is general because it offers the definition, identification, estimation, and sensitivity analysis of causal mediation effects without reference to any specific statistical model. Further, our approach explicitly links these 4 elements closely together within a single framework. As a result, the proposed framework can accommodate linear and nonlinear relationships, parametric and nonparametric models, continuous and discrete mediators, and various types of outcome variables. The general definition and identification result also allow us to develop sensitivity analysis in the context of commonly used models, which enables applied researchers to formally assess the robustness of their empirical conclusions to violations of the key assumption. We illustrate our approach by applying it to the Job Search Intervention Study. We also offer easy-to-use software that implements all our proposed methods. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

A self-agency bias in preschoolers' causal inferences.

Preschoolers' causal learning from intentional actions—causal interventions—is subject to a self-agency bias. The authors propose that this bias is evidence-based, in other words, that it is responsive to causal uncertainty. In the current studies, two causes (one child controlled, one experimenter controlled) were associated with one or two effects, first independently, then simultaneously. When initial independent effects were probabilistic, and thus subsequent simultaneous actions were causally ambiguous, children showed a self-agency bias. Children showed no bias when initial effects were deterministic. Further controls established that children's self-agency bias is not a wholesale preference but rather is influenced by uncertainty in causal evidence. These results demonstrate that children's own experience of action influences their causal learning, and the findings suggest possible benefits in uncertain and ambiguous everyday learning contexts. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Perception as evidence accumulation and Bayesian inference: Insights from masked priming.

The authors argue that perception is Bayesian inference based on accumulation of noisy evidence and that, in masked priming, the perceptual system is tricked into treating the prime and the target as a single object. Of the 2 algorithms considered for formalizing how the evidence sampled from a prime and target is combined, only 1 was shown to be consistent with the existing data from the visual word recognition literature. This algorithm was incorporated into the Bayesian Reader model (D. Norris, 2006), and its predictions were confirmed in 3 experiments. The experiments showed that the pattern of masked priming is not a fixed function of the relations between the prime and the target but can be changed radically by changing the task from lexical decision to a same-different judgment. Implications of the Bayesian framework of masked priming for unconscious cognition and visual masking are discussed. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Knowing who dunnit: Infants identify the causal agent in an unseen causal interaction.

Preverbal infants can represent the causal structure of events, including distinguishing the agentive and receptive roles and categorizing entities according to stable causal dispositions. This study investigated how infants combine these 2 kinds of causal inference. In Experiments 1 and 2, 9.5-month-olds used the position of a human hand or a novel puppet (causal agents), but not a toy train (an inert object), to predict the subsequent motion of a beanbag. Conversely, in Experiment 3, 10- and 7-month-olds used the motion of the beanbag to infer the position of a hand but not of a toy block. These data suggest that preverbal infants expect a causal agent as the source of motion of an inert object. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

The priority of just-identified, recursive models.

Theories often place constraints on causal relationships, and such constraints are often assessed with causal models. Causal models should be recursive and just identified because cause is recursive and is more likely to be just identified than overidentified. A just-identified, recursive model (JIRM) is specified that satisfies both requirements and that can be used to assess a wide range of causal implications in either a norm-referenced or criterion-referenced manner. P. E. Meehl and N. G. Waller (2002) proposed an innovative method for theory appraisal called the delete one-add one (D1 -A1) method, which assesses a relatively narrow range of causal implications, allows nonrecursive models, and is only norm referenced. The JIRM and D1-A1 methods are compared. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Causal knowledge and categories: The effects of causal beliefs on categorization, induction, and similarity.

Despite the recent interest in the theoretical knowledge embedded in human representations of categories, little research has systematically manipulated the structure of such knowledge. Across four experiments this study assessed the effects of interattribute causal laws on a number of category-based judgments. The authors found that (a) any attribute occupying a central position in a network of causal relationships comes to dominate category membership, (b) combinations of attribute values are important to category membership to the extent they jointly confirm or violate the causal laws, and (c) the presence of causal knowledge affects the induction of new properties to the category. These effects were a result of the causal laws, rather than the empirical correlations produced by those laws. Implications for the doctrine of psychological essentialism, similarity-based models of categorization, and the representation of causal knowledge are discussed. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Human judgments of positive and negative causal chains.

Three experiments investigated the way participants construct causal chains from experience with the individual links that make up those chains. Participants were presented with contingency information about the relationship between events A and B, as well as events B and C, using trial-by-trial presentations. The A-B and B-C contingencies could be positive, negative, or zero. Although participants had never experienced A and C together, A-C ratings were a multiplicative function of the A-B and B-C contingencies. These findings can be generated by an auto-associator using the delta rule. This explanation is also useful for understanding sensory preconditioning and second-order conditioning. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

How causal knowledge affects classification: A generative theory of categorization.

Several theories have been proposed regarding how causal relations among features of objects affect how those objects are classified. The assumptions of these theories were tested in 3 experiments that manipulated the causal knowledge associated with novel categories. There were 3 results. The 1st was a multiple cause effect in which a feature's importance increases with its number of causes. The 2nd was a coherence effect in which good category members are those whose features jointly corroborate the category's causal knowledge. These 2 effects can be accounted for by assuming that good category members are those likely to be generated by a category's causal laws. The 3rd result was a primary cause effect, in which primary causes are more important to category membership. This effect can also be explained by a generative account with an additional assumption: that categories often are perceived to have hidden generative causes. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Mechanisms of predictive and diagnostic causal induction.

In predictive causal inference, people reason from causes to effects, whereas in diagnostic inference, they reason from effects to causes. Independently of the causal structure of the events, the temporal structure of the information provided to a reasoner may vary (e.g., multiple events followed by a single event vs. a single event followed by multiple events). The authors report 5 experiments in which causal structure and temporal information were varied independently. Inferences were influenced by temporal structure but not by causal structure. The results are relevant to the evaluation of 2 current accounts of causal induction, the Rescorla-Wagner (R. A. Rescorla & A. R. Wagner, 1972) and causal model theories (M. R. Waldmann & K. J. Holyoak, 1992). (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Comparative outcome evaluations of psychotherapies: Guidelines for addressing eight limitations of the gold standard of causal inference.

Although randomized experiments represent the gold standard in causal research, their limitations are easily overlooked, especially in field experiments. Although discussions of the limitations of field experiments typically detail problems of design implementation, field experiments have limitations even if their randomized designs are successfully implemented. This is particularly true of comparative evaluations of psychotherapy programs. Accordingly, this article focuses on eight limitations that pervade such comparative treatment research: (a) Direct and indirect effects are conflated; (b) the investigation of interactions is limited; (c) attention is diverted from background conditions; (d) temporal factors complicate causal inferences; (e) the stable-unit-treatment-value assumption is easily violated; (f) therapist effects can confound therapy method effects; (g) differential effects of treatment methods are small; and (h) generalizability of treatment research findings is indeterminate. In addition to explaining each limitation, the author proposes research strategies for addressing it. The author concludes that attention to these 8 limitations and to the various strategies for overcoming them can increase increase the causal contribution of experimental evaluations of psychotherapies. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Causal inference and developmental psychology.

Causal inference is of central importance to developmental psychology. Many key questions in the field revolve around improving the lives of children and their families. These include identifying risk factors that if manipulated in some way would foster child development. Such a task inherently involves causal inference: One wants to know whether the risk factor actually causes outcomes. Random assignment is not possible in many instances, and for that reason, psychologists must rely on observational studies. Such studies identify associations, and causal interpretation of such associations requires additional assumptions. Research in developmental psychology generally has relied on various forms of linear regression, but this methodology has limitations for causal inference. Fortunately, methodological developments in various fields are providing new tools for causal inference—tools that rely on more plausible assumptions. This article describes the limitations of regression for causal inference and describes how new tools might offer better causal inference. This discussion highlights the importance of properly identifying covariates to include (and exclude) from the analysis. This discussion considers the directed acyclic graph for use in accomplishing this task. With the proper covariates having been chosen, many of the available methods rely on the assumption of “ignorability.” The article discusses the meaning of ignorability and considers alternatives to this assumption, such as instrumental variables estimation. Finally, the article considers the use of the tools discussed in the context of a specific research question, the effect of family structure on child development. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

A Theory of Causal Learning in Children: Causal Maps and Bayes Nets.

The authors outline a cognitive and computational account of causal learning in children. They propose that children use specialized cognitive systems that allow them to recover an accurate "causal map" of the world: an abstract, coherent, learned representation of the causal relations among events. This kind of knowledge can be perspicuously understood in terms of the formalism of directed graphical causal models, or Bayes nets. Children's causal learning and inference may involve computations similar to those for learning causal Bayes nets and for predicting with them. Experimental results suggest that 2- to 4-year-old children construct new causal maps and that their learning is consistent with the Bayes net formalism. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Limitations of exemplar models of multi-attribute probabilistic inference.

Observers were presented with pairs of objects varying along binary-valued attributes and learned to predict which member of each pair had a greater value on a continuously varying criterion variable. The predictions from exemplar models of categorization were contrasted with classic alternative models, including generalized versions of a "take-the-best" model and a weighted-additive model, by testing structures in which interactions between attributes predicted the magnitude of the criterion variable. Under typical training conditions, observers showed little sensitivity to the attribute interactions, thereby challenging the predictions from the exemplar models. In a condition involving highly extended training, observers eventually learned the relations between the attribute interactions and the criterion variable. However, an analysis of the observers' response times for making their paired-comparison decisions also challenged the exemplar model predictions. Instead, it appeared that most observers recoded the interacting attributes into emergent configural cues. They then applied a set of hierarchically organized rules based on the priority of the cues to make their decisions. (PsycINFO Database Record (c) 2010 APA, all rights reserved)