[2C3-OS-9a] 人工知能におけるプライバシー，公平性，説明責任，透明性への学際的アプローチ(1/2)

[2C3-OS-9a-01] Which are Differential Treatment Morally Bad?

Consideration of statistical discrimination in the area of insurance

〇Haruka Maeda^1,2,3 (1. The University of Tokyo Graduate School, 2. RIKEN Center for Advanced Intelligence Project, 3. Japan Society for the Promotion of Science)

Keywords:AI, discrimination, statistical discrimination

This presentation aims to show how to apply a way to analyze statistical discrimination in insurance to algorithmic discrimination. Recently algorithmic discrimination and technical fix are getting attention. Race or gender are well-known attributes that more likely to provoke discussion if there are disparities by such properties. However, it is not plausible to prohibit the use of such attributes and we do not know how to treat them if we do not use them. This presentation sheds light on when the differential treatment based on properties would ethically problematic by focusing on approaches of statistical discrimination in the area of insurance. These two approaches are based on likeliness without human’s malicious intension in common. More specifically, I will examine the possibility to apply through consideration of moral-related things in statistical in insurance. These considerations would offer a way to consider which case could be problematic.