Timezone: »

 
Spotlight
Learning Generalized Gumbel-max Causal Mechanisms
Guy Lorberbom · Daniel D. Johnson · Chris Maddison · Daniel Tarlow · Tamir Hazan

@

To perform counterfactual reasoning in Structural Causal Models (SCMs), one needs to know the causal mechanisms, which provide factorizations of conditional distributions into noise sources and deterministic functions mapping realizations of noise to samples. Unfortunately, the causal mechanism is not uniquely identified by data that can be gathered by observing and interacting with the world, so there remains the question of how to choose causal mechanisms. In recent work, Oberst & Sontag (2019) propose Gumbel-max SCMs, which use Gumbel-max reparameterizations as the causal mechanism due to an appealing counterfactual stability property. However, the justification requires appealing to intuition. In this work, we instead argue for choosing a causal mechanism that is best under a quantitative criteria such as minimizing variance when estimating counterfactual treatment effects. We propose a parameterized family of causal mechanisms that generalize Gumbel-max. We show that they can be trained to minimize counterfactual effect variance and other losses on a distribution of queries of interest, yielding lower variance estimates of counterfactual treatment effect than fixed alternatives, also generalizing to queries not seen at training time.

Author Information

Guy Lorberbom (Technion)
Daniel D. Johnson (Google Research, Brain Team)
Chris Maddison (University of Toronto)
Daniel Tarlow (Google Research, Brain team)
Tamir Hazan (Technion)

Related Events (a corresponding poster, oral, or spotlight)

More from the Same Authors