Methods of Designing Justice-oriented Interactive AI Systems

Abstract

The rise of artificial intelligence (AI) technologies has introduced concerns about the perpetuation and exacerbation of existing racial inequities. Although AI promises innovative advances, its design and implementation often reflect and amplify societal biases, disproportionately impacting marginalized groups, particularly Black individuals and groups of people. My dissertation explores this critical issue through examining how racial inequity manifests in two prominent interactive AI systems: banking applications and AI-enabled language tools. Existing research reveals shortcomings in current technology design practices, particularly in fostering accountability and inclusivity for underrepresented minority communities. While human-centered AI methods offer valuable tools, they may fall short in addressing the complex socio-cultural contexts of marginalized user groups. To bridge this gap, my dissertation contributes a provisional Techno-Realist Innovation Framework, a human-centered approach that integrates social justice principles into AI research and design. Through three distinct investigations, I explore the experiences of African American users with interactive AI systems, highlighting their unique challenges, benefits, and cultural assets. This research culminates in a reflective analysis and a methodological proposal for accountable community-based collaboration, emphasizing the importance of diverse community knowledge and participation in shaping technological innovation. By centering the voices and perspectives of marginalized groups, my dissertation seeks to pave the way for more just, equitable, and socially responsible AI systems.