A Decision-based View of Causality

By David E. Heckerman, Ross D. Shachter. 1994

Most traditional models of uncertainty have focused on the associational relationship among variables as captured by conditional dependence. In order to successfully manage intelligent systems for decision making, however, we must be able to predict the effects of actions. In this paper, we attempt to unite two branches of research that address such predictions: causal modeling and decision analysis. First, we provide a definition of causal dependence in decision-analytic terms, which we derive from consequences of causal dependence cited in the literature. Using this definition, we show how causal dependence can be represented within an influence diagram. In particular, we identify two inadequacies of an ordinary influence diagram as a representation for cause. We introduce a special class of influence diagrams, called causal influence diagrams, which corrects one of these problems, and identify situations where the other inadequacy can be eliminated. In addition, we describe the relationships between Howard Canonical Form and existing graphical representations of cause.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Explanation in Human-AI Systems: A Literature Meta-Review, Synopsis of Key Ideas and Publications, and Bibliography for Explainable AI (2019, background): “…Ahn and Kalish (2000); Chater and Oaksford, (2006); Cheng (1997, 2000); Einhorn and Hogarth (1986); Eiter and Lukasiewicz (2002, 2006); Feltovich, Hoffman, Woods, and Roesler (2004); Graesser, Baggett, and Williams (1996); Halpern and Pearl (2005a,b); Heckerman and Shachter (1994); Hesslow (1988).”
Learning Bayesian networks: The combination of knowledge and statistical data (1994, methodology): “In Heckerman et al. (1994), we describe a method for computing the cross entropy of two networks that makes use of the network structures. in our experiments, we construct prior networks by adding noise to the gold-standard network.”
Agent Incentives: A Causal Perspective (2021, background): “Similarly, the causality property for influence diagrams introduced by Heckerman and Shachter (1994) and Shachter and Heckerman (2010) only constrains the relationships to be partially causal downstream of the decision (though adding new decision-node parents to all nodes makes the diagram fully…”
A New Look at Causal Independence (1994, methodology): “Each node ≤σi is a dummy node that encodes the uncertainty in the relationships among eσi and its parents as described in Druzdel and Simon (1993) or, alternatively, Heckerman and Shachter (1994).”
Learning Bayesian networks (2007, methodology): “More recently, AI researchers and statisticians have begun to develop Bayesian methods for learning these networks from data Cooper and Herskovits, 1991, Buntine, 1991, Spiegelhalter et al., 1993, Heckerman et al., 1994, 1995].”
Causal independence for probability assessment and inference using Bayesian networks (1996, methodology): “In contrast, in a number of hardware troubleshooting applications [Heckerman et al., 1995a], we found the amechanistic form to be more e ective.”