GUARDRAILS

Two academic specialists look down the road at the evolution of AI and how to control it.

Discussions about how to regulate digital technology are inevitably heated and labyrinthine, with the participants often failing to agree on even the most basic of precepts. Anyone approaching this book with the expectation that Gasser, a professor in technology and social sciences at the Technical University of Munich, and Mayer-Schönberger, a professor of internet governance and regulation at Oxford, will lay out a one-size-fits-all model of regulation will be disappointed. Instead, the authors focus on establishing a conceptual framework that sets clear boundaries while still allowing for innovation and the capacity to change with dynamic circumstances. They use guardrails as an extended metaphor, looking at a wide array of cases, including the European Union’s attempts at tech regulation and the rules governing contributions to Wikipedia. Most attempts to date have shortcomings, but they provide lessons on how to balance competing concerns. Neither tech specialists nor ethicists can understand all the issues, and the proposals put forward by Gasser and Mayer-Schönberger involve collaboration and a willingness to compromise. The authors are wary of the “black boxes” in which AI systems operate, and they believe that handing decision-making power to machines is a dangerous path. An example of unintended consequences is the case where an algorithm rejected mortgage applications because they were made by Black applicants. The designers of algorithms must be able to explain exactly what is happening in an AI “box,” and algorithms need to be constructed to take into account social concerns, with ample provision for human oversight. Gasser and Mayer-Schönberger have interesting things to say about the topic, but the book is a dense, complex read, written with an academic audience in mind.

A scholarly framework for regulating AI technology, with an eye toward enhancing choice while promoting the social good.