In his January 2023 Communications Viewpoint, "The End of Programming," Matt Welsh wrote "nobody actually understands how large AI models work." However, already no one person understands existing large computer systems. Indeed, no team of people understands them. Staff turnover and other practicalities of real life mean not even the team that wrote them originally (should it still exist) nor the team currently responsible for maintaining them, fully understands large software systems, which can now exceed a billion lines of code. And yet such systems are in worldwide daily use and deliver economic benefits.

Rank aggregation is the problem of ordering a set of issues according to a set of individual rankings given as input. This problem has been studied extensively in computational social choice (see, e.g., Brandt et al. 2016) when the rankings are assumed to represent human preferences over, for example, candidates in a political election, projects to be funded, or more generally alternative proposals. The most common approach in this literature is to find normative desiderata for the aggregation process, including computational requirements such as the existence of tractable algorithms for its calculation and characterisations of the aggregators that satisfy them. Rank aggregation also has a wide spectrum of applications from metasearch engines [Dwork et al., 2001] to bioinformatics