Opening the government's black boxes: freedom of information and algorithmic accountability

Freedom of information laws are intended to illuminate how governments operate. However, the operations of governments increasingly involve algorithms, such as those used to recommend criminal sentencing and determine eligibility for social services. Algorithms function as black boxes' that turn inputs into outputs using processes that are often, by design, not transparent. Freedom of information laws allow one potential means for algorithmic transparency. However, whether such laws can be used to access algorithms is unclear. This research examines, in two ways, the availability of government algorithms to the public. First, this study examines laws, regulations, advisory opinions, and court rulings relevant to the disclosure of algorithms. The second part of this study analyzes actual responses by US government agencies to Freedom of Information Act requests for algorithms. This study concludes that governmental policies and practices related to algorithmic disclosure are inconsistent. Such inconsistencies suggest a need for better mechanisms to hold government algorithms accountable.