This course explores the interplay between biases, discrimination, and fairness in insurance, a field where the very logic of risk classification relies on differential treatment of policyholders. Insurance companies segment individuals into risk pools in order to assign premiums that reflect expected costs and incentivize risk reduction. While such segmentation is intrinsic to the industry, it raises critical questions: which forms of discrimination are permissible, where do biases in data and models intervene, and how do regulations around the world define acceptable practices? We will examine these questions by studying how discrimination may arise in predictive modeling for insurance, the measures used to detect it, and the methods available to mitigate unfairness—ranging from regulatory frameworks to statistical corrections. Lecture notes will be provided, based on selected chapters from Charpentier (2024).

On the mathematical side, the course emphasizes the role of modern quantitative tools in addressing fairness. After reviewing statistical and machine learning approaches to insurance pricing, we will introduce frameworks for assessing group fairness (such as demographic parity and equalized odds) and for analyzing counterfactual fairness, grounded in causal reasoning. A central mathematical focus will be on optimal transport and related concepts, which provide a versatile toolkit for measuring and mitigating disparities, both at the group and individual level. We will explore their use in post-processing adjustments of models, as well as in constructing counterfactuals for fairness evaluation.

References
Charpentier, A. (2024). Insurance: Biases, Discrimination, and Fairness. Springer. (Main course reference; lecture notes will be based on selected parts.)

Barocas, S., Hardt, M., & Narayanan, A. (2019). Fairness and Machine Learning. fairmlbook.org. (A standard reference on fairness definitions and methods in ML.)

Villani, C. (2009). Optimal Transport: Old and New. Springer. (Foundational text for mathematical background on optimal transport.)

Pearl, J., & Mackenzie, D. (2018). The Book of Why: The New Science of Cause and Effect. Basic Books. (Reference on causal inference and counterfactual reasoning, central to counterfactual fairness.)

要申込：講演を受講希望者は、Googleフォームにて申込みを行って下さい。QRコードまたは下記URLからアクセスしてください。
URL：https://forms.gle/3PEuSmPu2Bjy4dpA7

締切日： 10月27日 （月）

※数学・数理科学グローバル講義Ⅳは数学・数理科学イノベーション人材育成強化コースにおける
中核科目です。

※数学・数理科学グローバル講義Ⅳを履修するにはKULASIS での履修登録が必要です。
前期科目の履修登録期間は10 月10 日（金）～14 日（月）。

※履修登録していなくても聴講可（本学学生に限る。Googleフォーム申込みは必要）。

各特別講義のアブストラクト等の詳細は、当コースのホームページ
https://www.math.kyoto-u.ac.jp/ja/ktgu/courses　をご覧ください。