AI Ethics and Critique for Robotics

Abstract

In this thesis I consider using 3D computer vision for social good. In particu-lar, I present a broad and deep array of AI ethics methodologies and practices necessary to assess the harms and benefits of a particular AI technology or ap-plication. Many AI technologies are touted as being for social good, yet they are rightly critiqued as detached from the people and problems they purport to help and as ethics washing development and deployment of AI technolo-gies for surveillance or hegemonic interests. In this thesis I develop several AI ethics tools, including data and model audits and broader critiques andreimaginings of AI to understand the benefits and harms of AI. I also argue that the assessment of AI cannot happen isolated from impacted communities but rather must be led by them. To this end I present my work co-founding and leading Queer in AI, a community of LGBTQ+ AI practitioners working to combat bias and other AI harms towards queer people and imagine libra-tory queer AIs. I present my work on 3D computer vision, and throughout show how my ethics research shaped the technical problems I choose and ledme to identify new and socially beneficial problems in 3D computer vision.