Ralph Richard Banks (Stanford Law School) & Victor Y. Wu (Stanford University) have posted From Intent to Impact: Equal Protection in the Age of Artificial Intelligence on SSRN. Here is the abstract:
Artificial intelligence is rapidly reshaping governmental decision-making, permeating domains ranging from law enforcement and benefits administration to public health and education. Yet current equal protection doctrine remains anchored to notions of discriminatory intent and formal classification, frameworks that will become increasingly inapplicable in an algorithmic age. The opacity, complexity, and autonomy of machine learning models make it nearly impossible to attribute discriminatory purpose to the people who construct or rely on such models. At the same time, AI’s use of data so correlated with racial and other protected-class disparities means that those disparities will often persist even in the absence of reliance on formal classifications.
For these reasons, the transformation in governmental decision-making wrought by AI will place existing doctrine under tremendous pressure. The primacy of intent and classification will eventually diminish as the proliferation of AI reveals the incoherence and unworkability of those doctrinal criteria. Instead, a concern for impact and outcomes—historically incorporated into disparate impact law yet long marginalized in equal protection doctrine—will move to the fore. Contrary to prevailing assumptions, meaningful mitigation of algorithmic harms requires what scholars have termed “fairness through awareness,” not formal colorblindness.
Ultimately, AI may accomplish what decades of scholarly criticism have not: the abandonment of intent as a constitutional lodestar and an embrace of an impact-oriented framework more attuned to systemic inequality. Indeed, the doctrine’s reliance on intent and classification was already problematic even before the rise of AI. But as algorithmic decision-making becomes increasingly embedded in the machinery of government, the survival of the Equal Protection Clause as a meaningful constraint will force a reorientation toward an impact-oriented standard.