Beyond Algorithms: Toward a Normative Theory of Automated Regulation

Abstract

The proliferation of artificial intelligence in our daily lives has spawned a burgeoning literature on the dawn of dehumanized, algorithmic governance. Remarkably, the scholarly discourse overwhelmingly fails to acknowledge that automated, non-human governance has long been a reality. For more than a century, policy-makers have relied on regulations that automatically adjust to changing circumstances, without the need for human intervention. This Article surveys the track record of self-adjusting governance mechanisms to propose a normative theory of automated regulation. Effective policy-making frequently requires anticipation of future developments, from technology innovation to geopolitical change. Self-adjusting regulation offers an insurance policy against the well-documented inaccuracies of even the most expert forecasts, reducing the need for costly and time-consuming administrative proceedings. Careful analysis of empirical evidence, existing literature, and precedent reveals that the benefits of regulatory automation extend well beyond mitigating regulatory inertia. From a political economy perspective, automated regulation can accommodate a wide range of competing beliefs and assumptions about the future to serve as a catalyst for more consensual policy-making. Public choice theory suggests that the same innate diversity of potential outcomes makes regulatory automation a natural antidote to the domination of special interests in the policy-making process. Today’s automated regulations rely on relatively simplistic algebra, a far cry from the multivariate calculus behind smart algorithms. Harnessing the advanced mathematics and greater predictive powers of artificial intelligence could provide a significant upgrade for the next generation of automated regulation. Any gains in mathematical sophistication, however, will likely come at a cost if the widespread scholarly skepticism toward algorithmic governance is any indication of future backlash and litigation. Policy-makers should consider carefully whether their objectives may be served as well, if not better, through more simplistic, but well-established methods of regulatory automation.