As AI systems are deployed across an ever-growing number of domains, the fairness, transparency, and accountability of these systems has become a critical societal concern. This Program examines the intersections between AI and some of humanity’s most fundamental values, addressing urgent questions about algorithmic equity, explainability, responsibility, and inclusion.
Through original research and multistakeholder input, our Fairness, Transparency, and Accountability work asks how AI can build a world that is more (and not less) just than the one that came before it. And by offering actionable resources for implementing transparency at scale, ABOUT ML seeks to operationalize these insights with full-cycle documentation of machine learning systems.
Our Fairness, Transparency, and Accountability & ABOUT ML Work
Fairness, Transparency, and Accountability encompasses PAI’s large body of research and programming around algorithmic fairness, explainability, criminal justice, and diversity and inclusion. In 2020 alone, this work examined the challenges organizations face when they seek to measure and mitigate algorithmic bias using demographic data, provide meaningful explanations to diverse stakeholders, address bias in recidivism risk assessment tools, and build more inclusive AI teams.
With ABOUT ML, PAI is leading a multistakeholder effort to develop guidelines for the documentation of machine learning systems, setting new industry norms for transparency in AI. This means not just identifying the necessary components of transparency, but releasing actionable resources to help organizations operationalize transparency at scale. Developed through an iterative, multistakeholder process, these resources pool the collective efforts and insights of academic researchers, industry practitioners, civil society organizations, and the impacted public.
Updates & Media Coverage
Eyes Off My Data: Exploring Differentially Private Federated Statistics To Support Algorithmic Bias Assessments Across Demographic…
ABOUT ML in Practice: An Example from the Humanitarian Sector
Improving Documentation in Practice: Our First ABOUT ML Pilot
Shedding Light on the Trade-offs of Using Demographic Data for Algorithmic Fairness
Knowing the Risks: A Necessary Step to Using Demographic Data for Algorithmic Fairness
Crucial Yet Overlooked: Why We Must Reconcile Legal and Technical Approaches to Algorithmic Bias
From Affirmative Action to Affirmative Algorithms: The Legal Challenges Threatening Progress on Algorithmic Fairness
Why PATTERN Should Not Be Used: The Perils of Using Algorithmic Risk Assessment Tools During COVID-19