Purdue University

Amin Jaber, a PhD student in computer science working with the Regenstrief Center for Healthcare Engineering, won the Best Student Paper at the Conference on Uncertainty in Artificial Intelligence (UAI 2018), held in Monterey, California. The paper is titled “Causal Identification under Markov Equivalence”, and was co-authored by Jiji Zhang from the Philosophy Department at Lingnan University and Elias Bareinboim, assistant professor in Computer Science at Purdue University.

The Conference on Uncertainty in Artificial Intelligence is an international conference on research related to knowledge representation, learning, and reasoning in the presence of uncertainty. The award-wining student paper will appear in the Journal for Artificial Intelligence Research. In addition, Jaber and his co-authors were awarded a $1,000 prize.

The paper covers the problem of causal effect identification, which is concerned with determining whether a causal effect can be computed from a combination of observational data and substantive knowledge of the domain. Jaber and his co-authors combined the methods of causal graphs and Markov equivalence classes for causal identification.

The award was presented at the conference banquet on August 8th. Further information about UAI 2018 may be found at: http://auai.org/uai2018/schedule.php#schedule. The full paper may be downloaded from: https://www.dropbox.com/s/xt34vt0lkj3mto0/proceedings.pdf?dl=0.

ABSTRACT

Assessing the magnitude of cause-and-effect relations is one of the central challenges found throughout the empirical sciences. The problem of identification of causal effects is concerned with determining whether a causal effect can be computed from a combination of observational data and substantive knowledge about the domain under investigation, which is formally expressed in the form of a causal graph. In many practical settings, however, the knowledge available for the researcher is not strong enough so as to specify a unique causal graph. Another line of investigation attempts to use observational data to learn a qualitative description of the domain called a Markov equivalence class, which is the collection of causal graphs that share the same set of observed features. In this paper, we marry both approaches and study the problem of causal identification from an equivalence class, represented by a partial ancestral graph (PAG). We start by deriving a set of graphical properties of PAGs that are carried over to its induced subgraphs. We then develop an algorithm to compute the effect of an arbitrary set of variables on an arbitrary outcome set. We show that the algorithm is strictly more powerful than the current state of the art found in the literature.