Modeling Causal Reinforcement and Undermining with Noisy-AND Trees

Abstract. Causal modeling, such as noisy-OR, reduces probability parameters to be acquired in constructing a Bayesian network. Multiple causes can reinforce each other in producing the effect or can undermine the impact of each other. Most existing causal models do not consider their interactions from the perspective of reinforcement or undermining. We show that none of them can represent both interactions. We present the first explicit causal model that can encode both reinforcement and undermining and we show how to use such a model to support efficient probability elicitation.

• Type: Conference
• Year: 2006
• Where: AI
• Authors: Y. Xiang, N. Jia
• Keywords: AI 2006, Artificial Intelligence, Causal Model, Causal Modeling, Most Existing Causal