We present a framework for the rational analysis of elemental causal induction – learning about the existence of a relationship between a single cause and effect – based upon causal graphical models. This framework makes precise the distinction between causal structure and causal strength: the difference between asking whether a causal relationship exists and asking how strong that causal relationship might be. We show that two leading rational models of elemental causal induction, ∆P and causal power, both estimate causal strength, and introduce a new rational model, causal support, that assesses causal structure. Causal support predicts several key phenomena of causal induction that cannot be accounted for by other rational models, which we explore through a series of experiments. These phenomena include the complex interaction between ∆P and the base-rate probability of the effect in the absence of the cause, sample size effects, inferences from incomplete contingency tables, and causal learning from rates. Causal support also provides a better account of a number of existing datasets than either ∆P or causal power.

Inducing causal relationships from observations is a classic problem in scientific inference, statistics, and machine learning. It is also a central part of human learning, and a task that people perform remarkably well given its notorious difficulties. People can learn causal structure in various settings, from diverse forms of data: observations of the co-occurrence frequencies between causes and effects, interactions between physical objects, or patterns of spatial or temporal coincidence. These different modes of learning are typically thought of as distinct psychological processes and are rarely studied together, but at heart they present the same inductive challenge—identifying the unobservable mechanisms that generate observable relations between variables, objects, or events, given only sparse and limited data. We present a computational-level analysis of this inductive problem and a framework for its solution, which allows us to model all these forms of causal learning in a common language. In this framework, causal induction is the product of domain-general statistical inference guided by domain-specific prior knowledge, in the form of an abstract causal theory. We identify 3 key aspects of abstract prior knowledge—the ontology of entities, properties, and relations that organizes a domain; the plausibility of specific causal relationships; and the functional form of those relationships—and show how they provide the constraints that people need to induce useful causal models from sparse data.

Abstract. I argue that psychologists interested in human causal judgment should understand and adopt a representation of causal mechanisms by directed graphs that encode conditional independence (screening off) relations. I illustrate the benefits of that representation, now widely used in computer science and increasingly in statistics, by (i) showing that a dispute in psychology between ‘mechanist’ and ‘associationist ’ psychological theories of causation rests on a false and confused dichotomy; (ii) showing that a recent, much-cited experiment, purporting to show that human subjects, incorrectly let large causes ‘overshadow ’ small causes, misrepresents the most likely, and warranted, causal explanation available to the subjects, in the light of which their responses were normative; (iii) showing how a recent psychological theory (due to P. Cheng) of human judgment of causal power can be considerably generalized: and (iv) suggesting a range of possible experiments comparing human and computer abilities to extract causal information from associations.

The rationality of human causal judgments has been the focus of a great deal of recent research. We argue against two major trends in this research, and for a quite different way of thinking about causal mechanisms and probabilistic data. Our position rejects a false dichotomy between "mechanistic " and "probabilistic " analyses of causal inference-- a dichotomy that both overlooks the nature of the evidence that supports the induction of mechanisms and misses some important probabilistic implications of mechanisms. This dichotomy has obscured an alternative conception of causal learning: for discrete events, a central adaptive task is to induce causal mechanisms in the environment from probabilistic data and prior knowledge. Viewed from this perspective, it is apparent that the probabilistic norms assumed in the human causal judgment literature often do not map onto the mechanisms generating the probabilities. Our alternative conception of causal judgment is more congruent with both scientific uses of the notion of causation and observed causal judgments of untutored reasoners. We illustrate some of the relevant variables under this conception, using a framework for causal representation now widely adopted in computer science and, increasingly, in statistics. We also review the formulation and evidence for a theory of human causal induction (Cheng, 1997) that adopts this alternative conception. 1. The Old Mechanism Approach A long and still popular tradition in the study of human causal reasoning insists on a dramatic bifurcation between "mechanistic " conceptions of causalGlymour & Cheng inference and "probabilistic " or "covariational " conceptions of this process (e.g., Ahn

People routinely make sophisticated causal inferences unconsciously, effortlessly, and from very little data – often from just one or a few observations. We argue that these inferences can be explained as Bayesian computations over a hypothesis space of causal graphical models, shaped by strong top-down prior knowledge in the form of intuitive theories. We present two case studies of our approach, including quantitative models of human causal judgments and brief comparisons with traditional bottom-up models of inference. 1

Abstract not found

Current psychological theories of human causal learning and judgment focus primarily on long-run predictions: two by estimating parameters of a causal Bayes nets (though for different parameterizations), and a third through structural learning. This paper focuses on people’s short-run behavior by examining dynamical versions of these three theories, and comparing their predictions to a real-world dataset. 1

Much research on judgment and decision making has focussed on the adequacy of classical rationality as a description of human reasoning. But more recently it has been argued that classical rationality should also be rejected even as normative standards for human reasoning. For example, Gigerenzer and Goldstein (1996) and Gigerenzer and Todd (1999a) argue that reasoning involves ‘‘fast and frugal’ ’ algorithms which are not justified by rational norms, but which succeed in the environment. They provide three lines of argument for this view, based on: (A) the importance of the environment; (B) the existence of cognitive limitations; and (C) the fact that an algorithm with no apparent rational basis, Take-the-Best, succeeds in an judgment task (judging which of two cities is the larger, based on lists of features of each city). We reconsider (A)–(C), arguing that standard patterns of explanation in psychology and the social and biological sciences, use rational norms to explain why simple cognitive algorithms can succeed. We also present new computer simulations that compare Take-the-Best with other cognitive models (which use connectionist, exemplarbased, and decision-tree algorithms). Although Take-the-Best still performs well, it does not perform noticeably better than the other models. We conclude that these results provide no strong reason to prefer Take-the-Best over alternative cognitive models.

Major biases and stereotypes in group judgments are reviewed and modeled from a recurrent connectionist perspective. These biases are in the areas of group impression formation (illusory correlation), group differentiation (accentuation), stereotype change (dispersed vs. concentrated distribution of inconsistent information), and group homogeneity. All these phenomena are illustrated with well-known experiments, and simulated with an autoassociative network architecture with linear activation update and delta learning algorithm for adjusting the connection weights. All the biases were successfully reproduced in the simulations. The discussion centers on how the particular simulation specifications compare with other models of group biases and how they may be used to develop novel hypotheses for testing the connectionist modeling approach and, more generally, for improving theorizing in the field of social biases and stereotype change. Petite, attractive, intelligent, WSF, 30, fond of music, theatre, books, travel, seeks warm, affectionate, fun-loving man to share life’s pleasures with view to lasting relationship. Send photograph. Please no

How is information transmitted in a group? A multi-agent connectionist model is proposed that combines features of standard recurrent models to simulate the process of information uptake, integration and memorization within individual agents, with novel aspects that simulate the communication of beliefs and opinions between agents. A crucial aspect in belief updating based on information from other agents is the trust in the information provided, implemented as the consistency with the receiving agents’ existing beliefs. Trust leads to a selective propagation and thus filtering out of less reliable information, and implements Grice’s (1975) maxims of quality and quantity in communication. By studying these communicative aspects within the framework of standard models of information processing, the unique contribution of communicative mechanisms beyond intra-personal factors was explored in simulations of key phenomena involving persuasive communication and polarization, lexical acquisition, spreading of stereotypes and rumors, and a lack of sharing unique information in group decisions.