Advancing social justice through linguistic justice: Strategies for building equity fluent NLP technology
Julia Nee, Genevieve Macfarlane Smith, Alicia Myles Sheares, Ishita Rustagi · 2021
Language and social reality are mutually reinforcing; as a result, natural language processing (NLP) presents a unique opportunity to shift social reality at scale, advancing social justice by promoting linguistic justice. We provide an overview of how language and bias are intertwined and implications for building NLP tools that actively advance equity and inclusion. Then, we present a framework for centering inclusion and social justice in NLP design at four overlapping layers of linguistic structure. The goal is to provide a foundation for adopting equity-centered principles in the creation of NLP tools that don’t simply mitigate social biases, but actively advance inclusion and social justice through language. This work aims to be practical and builds from a partnership between researchers at the Center for Equity, Gender, and Leadership at the UC Berkeley Haas School of Business and leaders and practitioners at a large Silicon Valley tech firm. This framework can foster equity-centered thinking to lead to greater “equity fluent” NLP tools that have the potential to advance justice more broadly.