Algorithmic Bias and the New Chicago School

Algorithmic bias stems from two primary sources: the training data fed into AI systems and the design of the algorithms themselves, meaning bias can be embedded at multiple stages of AI development, not just in how data is collected.

Transparency and explanation requirements alone are insufficient: While many legal scholars advocate for direct regulation through transparency obligations, this approach alone is incomplete and must be paired with other regulatory strategies to effectively address algorithmic bias.

A multi-layered regulatory approach is more effective than single solutions: The paper proposes combining direct regulation (through law) with indirect regulation across four dimensions—architecture (technical design), norms (social standards), market forces (economic incentives), and legal frameworks—to create comprehensive protection against biased AI decisions.

Current AI deployment poses significant societal risks: AI systems making independent decisions in both public and private sectors can have "far-reaching impact on individuals and society," suggesting that algorithmic bias is not merely a technical problem but a critical governance challenge requiring coordinated solutions.

Regulation should address both how AI is built and how it operates: Effective oversight requires attention to both the structural design of algorithms and the broader ecosystem (market incentives, social norms, technical architecture) in which they function, rather than relying solely on post-deployment transparency measures.