Brief Introduction to Causal Compositional Models
When applying probabilistic models to support decision making processes, the users have to strictly distinguish whether the impact of their decision changes the considered situation or not. In the former case it means that they are planing to make an intervention, and its respective impact cannot be estimated from a usual stochastic model but one has to use a causal model. The present paper thoroughly explains the difference between conditioning, which can be computed from both usual stochastic model and a causal model, and computing the effect of intervention, which can only be computed from a causal model. In the paper a new type of causal models, so called compositional causal models are introduced. Its great advantage is that both conditioning and the result of intervention are computed in very similar ways in these models. On an example, the paper illustrates that like in Pearl’s causal networks, also in the described compositional models one can consider models with hidden variables.
KeywordsBayesian Network Causal Relation Conditional Distribution Causal Model Hide Variable
This work was supported in part by the National Science Foundation of the Czech Republic by grant no. GACR 15-00215S.
- 1.Bína, V., Jiroušek, R.: Marginalization in multidimensional compositional models. Kybernetika 42(4), 405–422 (2006)Google Scholar
- 5.Hagmayer, Y., Sloman, S., Lagnado, D., Waldmann, M.R.: Causal reasoning through intervention. In: Gopnik, A., Schulz, L. (eds.) Causal Learning: Psychology, Philosophy, and Computation, pp. 86–101. Oxford University Press, Oxford (2002)Google Scholar
- 8.Jiroušek, R.: Brief introduction to probabilistic compositional models. Uncertainty analusis in econometrics with applications. In: Huynh, V.N., Kreinovich, V., Sriboonchita, S., Suriya, K. (eds.) AISC 200, pp. 49–60. Springer, Berlin (2013)Google Scholar
- 9.Jiroušek, R.: On causal compositional models: simple examples. In: Laurent, A. et al. (eds.) Proceedings of the 15th International Conference on Information Processing and Management of Uncertainty in Knowledge-Based Systems. Part I, CCIS 442, Springer International Publishing, Switzerland, pp. 517–526 (2014)Google Scholar
- 12.Jiroušek, R., Vejnarová, J., Daniel, M.: Compositional models of belief functions. In: de Cooman, G., Vejnarová, J., Zaffalon, M. (eds.) Proceedings of the Fifth International Symposium on Imprecise Probability: Theories and Applications, Praha, pp. 243–252 (2007)Google Scholar
- 13.Lauritzen, S.L.: Graphical Models. Oxford University Press, Oxford (1996)Google Scholar
- 16.Pearl, J.: Causality: Models, Reasoning, and Inference, Second Edition. Cambridge University Press, Cambridge (2009)Google Scholar
- 17.Ryall, M., Bramson, A.: Inference and Intervention: Causal Models for Business Analysis. Routledge, New York (2013)Google Scholar
- 20.Tucci, R.R.: Introduction to Judea Pearl’s Do-Calculus (2013). arXiv:1305.5506v1 [cs.AI]
- 21.Vejnarová, J.: Composition of possibility measures on finite spaces: preliminary results. In: Bouchon-Meunier, B., Yager, R.R. (eds.) Proceedings of 7th International Conference on Information Processing and Management of Uncertainty in Knowledge-based Systems IPMU’98, Editions E.D.K. Paris, pp. 25–30 (1998)Google Scholar