Repairing AI: How Breakages of Algorithmic Systems Reveal Repair Efforts and Possibilities for Renewal

Abstract: With the election of Donald Trump, many regulations for ethical, fair, and safe AI in the United States have been removed, and techno-optimistic rhetoric reigns over concerns for consumer safety and human rights. The European Union has been called out for stifling innovation through regulation. As the AI Act comes into effect, criticism of its many loopholes prevails, and oversight and enforcement have yet to materialize. But regulation is only one of many approaches to responsible AI. Direct responses to breaking, malfunctioning, or faulty technology allow for repair, shaping good practices, and informing good governance. This article explores the notions of breakage, repair, and renewal. Algorithmic systems are deployed in today's digital societies for various purposes but often appear unfit to perform their tasks satisfactorily. Compensating for their shortcomings through various forms of repair work, such as design changes and policy decisions, is essential. The notions of breakage, repair, and renewal allow for a more nuanced understanding of issues that arise from the promotion and implementation of AI systems, as well as a historically informed perspective on technology development and innovation.