開催日：2022年3月4日

会場：オンライン

京大の大塚淳です。滋賀大学の清水昌平先生より以下の集会について周知を依頼されましたので投稿します。よろしくおねがいします。
—–

滋賀大学の清水昌平です。因果と哲学に関する研究集会 International Workshop on Causality and Philosophy 2022のご案内です。どうぞよろしくお願いいたします。

International Workshop on Causality and Philosophy
日時: 2022年3月4日 16:00-18:00
プログラムと参加登録: https://www.ds.shiga-u.ac.jp/iwcp2022/
参加費: 無料
主催: 日本行動計量学会・滋賀大学データサイエンス教育研究センター
後援: 理研AIP

Invited speakers:
Jiji Zhang (Hong Kong Baptist University)
Title: Modularity and Causal Reasoning: A New Perspective
Abstract: Modularity plays a foundational role in graphical causal modeling. The idea that a causal system can be decomposed into a set of local modules underlies, on the one hand, the basic principles for reasoning about interventions such as the celebrated do-calculus, and　on the other hand, some novel techniques for learning causal structures from data. In this talk, we propose a new account of modularity that is at once more abstract and more intuitive, using the graphical language of string diagrams in category theory. We present a characterization of this formal notion of modularity in terms of graphical conditions on the causal structure, which unifies the standard rules of reasoning about interventions and is potentially useful for extending modularity-based causal discovery algorithms.

Jun Otsuka (Kyoto University & RIKEN)
Title: Three ways of modeling causality
Abstract: In the current mainstream statistics/machine learning
literature, causation is understood as a sort of relationship between variables, represented by a directed graph. In this talk we propose two alternative ways of conceptualizing causal relationships, with corresponding mathematical formulations. The first alternative is to conceptualize a causal structure as a system of connected mechanisms, where each mechanism passes its products to others. This analogy is best captured by the category-theoretic formulation where a causal model is represented as a functor between monoidal categories (Jacobs et al. 2019, Otsuka & Saigo, 2022). Secondly, a causal model can be understood as a set of laws that rule intervention calculus (e.g. do-calculus), which is a mapping of a distribution to another given an intervention, or monoid actions on the set of probability distributions. We argue that these three conceptualizations of causality are formally equivalent, but give different perspectives on the nature of causal relationships. This is joint work with Hayato Saigo.

Konstantin Genin (University of Tübingen)
Title: Success Concepts for Causal Discovery
Abstract: Existing causal discovery algorithms are often evaluated using two success criteria, one that is too strong to be feasible and the other which is too weak to be satisfactory. The unachievable criterion—uniform consistency—requires that a discovery algorithm identify the correct causal structure at a known sample size. The weak but achievable criterion—pointwise consistency—requires only that one identify the correct causal structure in the limit. We investigate two intermediate success criteria—decidability and progressive solvability—that are stricter than mere consistency but
weaker than uniform consistency. To do so, we review several topological theorems characterizing the causal discovery problems that are decidable and progressively solvable. We show, under a variety of common modeling assumptions, that there is no uniformly consistent procedure for identifying the direction of a causal edge, but there are statistical decision procedures and progressive solutions. We focus on linear models in which the error terms are either non-Gaussian or contain no Gaussian components, where the latter is relatively novel to the causal discovery literature. We focus especially on which success criteria remain feasible when confounders are present.

タグ：Causality