Intersectional AI

• DOI: 10.1145/3415218
• Publication Date: 2020-10-15T22:27:49Z
• Authors: Nora McDonald; Shimei Pan
• Keywords: Ethics; Algorithm bias; Artificial intelligence; 0508 media and communications; 4. Education; 05 social sciences; Surveillance technologies; 16. Peace & justice; Human-computer interaction; Risk assessment

Recent literature has demonstrated the limited and, in some instances, waning role of ethical training computing classes US. The capacity for artificial intelligence (AI) to be inequitable or harmful is well documented, yet it's an issue that continues lack apparent urgency effective mitigation. question we raise this paper how prepare future generations recognize and grapple with concerns a range issues plaguing AI, particularly when they are combined surveillance technologies ways have grave implications social participation restriction?from risk assessment bail assignment criminal justice, public benefits distribution access housing other critical resources enable security success within society. US mecca information computer science (IS CS) learning Asian students whose experiences as minorities renders them familiar with, vulnerable to, societal bias feeds AI bias. Our goal was better understand who being educated design systems think about these issues, particular, their sensitivity intersectional considerations heighten groups. In report on findings from qualitative interviews 20 graduate students, 11 class 9 Data Mining class. We find not predisposed deeply privacy well-being others unless explicitly encouraged do so. When do, thinking focused through lens personal identity experience, but reflections tend center bias, intrinsic feature design, rather than fairness, outcome requires imagine consequences AI. While are, fact, equipped fairness prompted by discussion exercises invite consideration intersectionality structural inequalities, many need help empathy 'work.' Notably, more frequently reflect problems related also likely consider connection between model attributes interaction context. suggest experience identity-based vulnerability promotes analytically complex lending further support argument identity-related ethics should integrated into IS CS curriculums, positioned stand-alone course.