State of the Evidence: Algorithmic Transparency

This paper presents a literature review on the implementation of algorithmic transparency in the public sector, examining the existing evidence on mechanisms to promote algorithmic transparency and their impacts and outcomes. To do this, we conducted keyword searches in databases of scientific publications and search engines, citations chaining, and followed expert recommendations in order to identify key documents in the area. A corpus of 41 relevant documents was reached, with information from 2015 to 2022 in English and Spanish. From the systematic review of these documents, multiple mechanisms to promote algorithmic transparency were identified and grouped into (1) disclosure of information, (2) explanations and (3) evaluations. These actions vary in terms of who makes the account of the algorithmic systems, in which stages of the algorithmic life cycle, and the actors or forums to which such actions are oriented. However, certain ambiguity remains regarding the definition, dimensions, implementation, and impacts of algorithmic transparency. It was also found that the outcomes of algorithmic transparency remain understudied, even though preliminary studies on the benefits and limits of algorithmic transparency were found based on scenario experiments and qualitative studies. Finally, the limitations of this research are discussed, such as the breadth of the search method that included very general and English-centered concepts, which could be taken into account for future research.

Keywords: Algorithmic transparency, Public Sector, AI, Accountability.

This document was used as the basis for OGP’s The Skeptic’s Guide to Open Government : Chapter 8 https://www.opengovpartnership.org/skeptics-guide-to-open-government-2022-edition/#toc_8