Biden should double down on Trump’s policy of promoting AI within government

President Biden is signing a flurry of executive actions during his first few weeks in office, many of them overturning Trump policies. Trump’s recent executive order on promoting the use of artificial intelligence in government agencies, however, presents a rare bipartisan sentiment and promises to improve government policies and services across the board. The current administration should not only maintain this policy, it should make it a priority.

AI is fundamentally changing the way people engage with technology, and the advantages of AI — enhanced problem solving and pattern detection, autonomously operated machines, and so on — extend beyond the private sector. AI can help governments produce informed and effective policy, optimize processes, improve quality of services, and engage the public.

The pandemic has made this even clearer. Many government agencies rely on AI to monitor and treat COVID. The Pentagon is using AI to predict water, medicine, and supply shortages. AI is even helping the Department of Energy identify molecules to test in the lab as potential COVID treatments. The list goes on.

AI also has its uses outside of the pandemic. For example, Pittsburgh used AI to cut down on traffic, reducing travel times by 25% and cutting emissions. Chicago is even using AI to prevent crimes by predicting when and where they are likely to happen (no Precogs needed). Unfortunately, pre-COVID examples like these are scarce.

This is why the previous administration’s executive order is important. It lays out a plan for establishing government-wide guidance (or standards) for the adoption of AI within federal agencies. This represents a break from traditional policy, which lumped AI in with other technologies, relying on old standards for a new, invariably different technology.

The Biden administration should prioritize this policy because standards set the tone for agency staff. Standards not only indicate that the use of certain tools is permitted or encouraged, but they also provide the roadmaps necessary for staff to feel comfortable adopting the technology — and to do so effectively. Creating a common set of standards across all agencies will improve information sharing between agencies, too. This will conceivably increase agency effectiveness and efficiency, as well as the quality of the standards.

Consider a government-wide AI standard for public documents. This standard might require agencies to make all public documents machine readable and include tags, allowing users to quickly and easily sort by topic, search for keywords, or aggregate data from related documents. This would open up large swaths of data for use by the public and private sectors. For example, business owners could quickly identify regulations that apply to their businesses without having to read the nearly 200,000 pages of regulations currently on the books. And that’s just one of countless possibilities for new standards.

Another reason to prioritize this policy is that it promotes goals outlined in the Biden administration’s executive actions. The goal of one such action is “to make evidence-based decisions guided by the best available science and data.” In today’s day and age, AI is often the best scientific tool and provides the best data.

This Biden action also instructs agencies to “expand open and secure access to Federal data,” including a mandate to make collected data available in machine-readable format. This is already one step toward common AI standards across agencies. The administration could even take this a step further by extending it to all documents.

In another executive order, the Biden administration states its goal to produce “a set of recommendations for improving and modernizing regulatory review.” With hundreds of thousands of regulatory pages on the books, one way to modernize review would be to begin with outdated or duplicative regulations, which agencies can identify using AI (a recommendation from the previous administration’s AI policy).

AI also presents new ways to evaluate the success and broader effects of existing regulations. Agencies could use this knowledge to inform a modernized regulatory review process and develop more effective regulations.

Expanding the use of AI in government is bipartisan policy — while people on either side of the aisle may prefer more or fewer policies in any specific area, everyone wants policies to be more informed and better constructed. However, there is a danger that the policy on AI standards for government gets lumped in with the Trump administration’s policy on AI regulations and standards for the private sector.

The private sector AI policy is not as bipartisan and has its critics. In fact, we responded to the original plan back in 2019, arguing that instead of developing standards for the private sector, the administration should turn its focus to government-wide standards for federal agencies. This is just what the previous administration did with the new AI policy.

Developing government-wide standards specific to AI will promote more and better use of AI within government agencies, leading to higher-quality policies and services. Those in the new administration should not let the origin of this policy blind them to its benefits and bipartisan nature. The Biden administration can use this opportunity to prioritize an effort that both parties can agree on — an effort that will expand the scientific grounding of government policies. That’s an effort that will mean a more effective government, as Biden would say, “for all Americans.”

Patrick McLaughlin is the director of policy analytics at the Mercatus Center at George Mason University, where he created and leads the RegData and QuantGov projects, deploying machine learning and other data-science tools to quantify governance indicators found in federal and state regulations and other policy documents.

Tyler Richards is the research coordinator of policy analytics at the Mercatus Center.