b'ViewpointsAlgorithmic Auditing: The Key to Making Machine Learning in the Public InterestBy Sara KassirAs the burdens for collecting enormous amounts of dataalso poses significant risks, particularly when algorithms decreased in recent years, advanced methods of analyzingare assumed to be infallible. While it may be true that this information rapidly developed. Machine learning, orthese applications process any information they are given the automation of model building, is one such method thatobjectively, human-generated data invariably reflects quickly became ubiquitous and impactful across industries.human biases. Therefore, automated tools can end up For the public sector, artificially intelligent algorithms areentrenching problematic simplifications about the world. now being deployed to solve problems that were previouslyBoth government and private sector industry players have viewed as insurmountable by humans. In internationalexperienced calls to proactively address this issue.development, they are working to predict areas susceptible to famine; in regulation, they are detecting the sources ofThis article argues that the risks of machine learning foodborne illness; in medicine, they are adding greater speedapplications are best mitigated through a process of and precision to diagnostic processes. algorithmic auditing, which institutionalizes accountability and robust due diligence in the technology. By assessing The advancements presented by big data and machinethe ways in which bias might emerge at each step in the learning are undeniably promising, but the technologydevelopment pipeline, it is possible to develop strategies 88 www.businessofgovernment.org The Business of Government'