Directly Discriminatory Algorithms

There is a broad consensus that algorithmic systems should be approached principally through the lens of indirect discrimination and their impact, but researchers argue that this approach is both normatively undesirable and legally flawed. In some cases, algorithmic bias may constitute direct discrimination, and the law should be better equipped to deal with this.

Empirical evidence of bias in automated decision-making will require us to revisit many of the fundamental challenges to the conceptual apparatus of discrimination law – from academic debates about the discipline's normative foundations to courts’ reluctance to approach causation through a ‘legalistic’ lens in the equality context.