ABSTRACT
With more and more governments deploying AI, responsible AI governance and algorithmic transparency are crucial to maintain public trust. One strategy to provide more transparency on government AI use is a public algorithm register, or a freely accessible database of algorithms used in the public sector with information on each algorithm’s scope, purpose, and function. Some public institutions have started experimenting with public algorithm registers, but they have room for improvement. In this chapter, we first summarize existing sets of recommendations and best practices for public algorithm registers. Then, based on the literature, our analysis of existing registers, and conversations with public servants, we present several recommendations to improve public algorithm registers. We focus in particular on the quality and quantity of information, the experience of citizens using the registers, accessibility for non-experts, and accountability. Public administrations must ensure that these registers also offer meaningful transparency if they want to reinforce public trust. Finally, we outline a number of key tensions in decision-making on algorithm registers and conclude that algorithm registers are just one piece of the puzzle, and they must be embedded in a broader digital and data literacy strategy.
