This paper describes a Bayesian method for combining an arbitrary mixture of observational and experimental data in order to learn causal Bayesian networks. Observational data are passively observed. Experimental data, such as that produced by randomized controlled trials, result from the experimenter manipulating one or more variables (typically randomly) and observing the states of other variables. The paper presents a Bayesian method for learning the causal structure and parameters of the underlying causal process that is generating the data, given that (1) the data contains a mixture of observational and experimental case records, and (2) the causal process is modeled as a causal Bayesian network. This learning method was applied using as input various mixtures of experimental and observational data that were generated from the ALARM causal Bayesian network. In these experiments, the absolute and relative quantities of experimental and observational data were varied systematically. For each of these training datasets, the learning method was applied to predict the causal structure and to estimate the causal parameters that exist among randomly selected pairs of nodes in ALARM that are not confounded. The paper reports how these structure predictions and parameter estimates compare with the true causal structures and parameters as given by the ALARM network.
Explaining why events occur is key to making decisions, assigning blame, and enacting policies. Despite the need, few methods can compute explanations in an automated way. Existing solutions start with a type-level model (e.g. factors affecting risk of disease), and use this to explain token-level events (e.g. cause of an individual's illness). This is limiting, since an individual's illness may be due to a previously unknown drug interaction. We propose a hybrid method for token explanation that uses known type-level models while also discovering potentially novel explanations. On simulated data with ground truth, the approach finds accurate explanations when observations match what is known, and correctly finds novel relationships when they do not. On real world data, our approach finds explanations consistent with intuition.
Reinforcement Learning (RL) is a heuristic method for learning locally optimal policies in Markov Decision Processes (MDP). Its classical formulation (Sutton & Barto 1998) maintains point estimates of the expected values of states or state-action pairs. Bayesian RL (Dearden, Friedman, & Russell 1998) extends this to beliefs over values. However the concept of values sits uneasily with the original notion of Bayesian Networks (BNs), which were defined (Pearl 1988) as having explicitly causal semantics. In this paper we show how Bayesian RL can be cast in an explicitly Bayesian Network formalism, making use of backwards-in-time causality. We show how the heuristic used by RL can be seen as an instance of a more general BN inference heuristic, which cuts causal links in the network and replaces them with noncausal approximate hashing links for speed. This view brings RL into line with standard Bayesian AI concepts, and suggests similar hashing heuristics for other general inference tasks.
We present a new approach to token-level causal reasoning that we call Sequences Of Mechanisms (SoMs), which models causality as a dynamic sequence of active mechanisms that chain together to propagate causal influence through time. We motivate this approach by using examples from AI and robotics and show why existing approaches are inadequate. We present an algorithm for causal reasoning based on SoMs, which takes as input a knowledge base of first-order mechanisms and a set of observations, and it hypothesizes which mechanisms are active at what time. We show empirically that our algorithm produces plausible causal explanations of simulated observations generated from a causal model. We argue that the SoMs approach is qualitatively closer to the human causal reasoning process, for example, it will only include relevant variables in explanations. We present new insights about causal reasoning that become apparent with this view. One such insight is that observation and manipulation do not commute in causal models, a fact which we show to be a generalization of the Equilibration-Manipulation Commutability of [Dash(2005)].
A major inference task in Bayesian networks is explaining why some variables are observed in their particular states using a set of target variables. Existing methods for solving this problem often generate explanations that are either too simple (underspecified) or too complex (overspecified). In this paper, we introduce a method called Most Relevant Explanation (MRE) which finds a partial instantiation of the target variables that maximizes the generalized Bayes factor (GBF) as the best explanation for the given evidence. Our study shows that GBF has several theoretical properties that enable MRE to automatically identify the most relevant target variables in forming its explanation. In particular, conditional Bayes factor (CBF), defined as the GBF of a new explanation conditioned on an existing explanation, provides a soft measure on the degree of relevance of the variables in the new explanation in explaining the evidence given the existing explanation. As a result, MRE is able to automatically prune less relevant variables from its explanation. We also show that CBF is able to capture well the explaining-away phenomenon that is often represented in Bayesian networks. Moreover, we define two dominance relations between the candidate solutions and use the relations to generalize MRE to find a set of top explanations that is both diverse and representative. Case studies on several benchmark diagnostic Bayesian networks show that MRE is often able to find explanatory hypotheses that are not only precise but also concise.