Public opinion lessons for AI regulation
This report from The Brookings Institution’s Artificial Intelligence and Emerging Technology (AIET) Initiative is part of “AI Governance,” a series that identifies key governance and norm issues related to AI and proposes policy remedies to address the complex challenges associated with emerging technologies.
An overwhelming majority of the American public believes that artificial intelligence (AI) should be carefully managed. Nevertheless, as the three case studies in this brief show, the public does not agree on the proper regulation of AI applications. Indeed, population-level support of an AI application may belie opposition by some subpopulations. Many AI applications, such as facial recognition technology, could cause disparate harm to already vulnerable subgroups, particularly ethnic minorities and low-income individuals. In addition, partisan divisions are likely to prevent government regulation of AI applications that could be used to influence electoral politics. In particular, the regulation of content recommendation algorithms used by social media platforms has been highly contestable. Finally, mobilizing an influential group of political actors, such as machine learning researchers in the campaign against lethal autonomous weapons, may be more effective in shifting policy debates than mobilizing the public at large.
AI is a general-purpose technology that enables machines to perform tasks that previously required human intelligence. As a result, it has a wide range of commercial and security applications. A 2018 survey conducted by the Center for the Governance of AI found that 84% of the American public believes that AI is a technology that should be carefully managed. Furthermore, the survey suggests that Americans consider most AI governance challenges to have high issue importance, as seen in Figure 1 below.
This brief focuses on how public opinion will likely shape the regulation of three applications of AI in the U.S.: facial recognition technology used by law enforcement, algorithms used by social media platforms, and lethal autonomous weapons. These case studies were selected because they involve AI governance issues that the American public characterize as either highly likely to impact them in the next decade or important for tech companies and governments to manage. Political debates around these applications touch on central themes articulated in numerous AI ethics principles, including fairness, privacy, and safety. As shown in the figure below, Americans predict some of these governance challenges as more likely to impact Americans in the next decade than others. The issues thought to be the most likely to impact Americans and rated the highest in issue importance include preventing AI-assisted surveillance from violating privacy and civil liberties, preventing AI from being used to spread fake and harmful content online, preventing AI cyberattacks, and protecting data privacy.
Facial recognition algorithms use facial features to identify, verify, and classify persons. The technology has widespread consumer applications, such as categorizing photos and unlocking smartphones. Law enforcement agencies have also begun to use facial recognition technology to scan through driver’s license photos and mug shots. The U.S. Customs and Border Protection is using facial recognition technology to screen international passengers at major airports.
Civil rights groups and academic researchers have criticized law enforcement’s adoption of facial recognition technology by citing concerns of racial and gender bias as well as a violation of civil rights. Researchers found that leading commercial facial recognition software programs are much less accurate at identifying women, particularly those with darker skin, than white men.
