Kevin Frazier is an Assistant Professor at the Crump College of Law at St. Thomas University. He previously clerked for the Montana Supreme Court.

Elections are often described as being “referendums” on recent policy decisions. The 2010 midterm supposedly signaled the public’s views on the Affordable Care Act. Similarly, the 2006 midterm theoretically amounted to a vote on the Iraq War. If Congress rushes to regulate artificial intelligence, then the upcoming election could, in part, be a proxy election on that AI policy.

Given the potential of AI to upend our economy, alter our culture, and hinder our democracy, why not actually put the topic of AI on the ballot?

The stakes are simply too high to only give the people an indirect vote on what may be the most consequential regulatory challenge yet to face the United States. Now’s the time for the Biden Administration and a multitude of U.S. representatives and senators to make good on their commitment to shape AI policy in response to the will of the American people. The best way for them to practice what they preach is to hold a national advisory referendum on when and how to regulate AI.

Let’s get some important questions out of the way. Would an AI referendum be legal? Yes, Congress can pass a statute to place a non-binding advisory question on the ballot. Pursuant to the "necessary and proper clause" or, as the founder's called it, "the sweeping clause," Congress has the authority to exercise all implied and incidental powers "conducive" to the "beneficial exercise" of one its enumerated powers, such as the regulation of interstate commerce and the promotion of the general welfare. A nonbinding referendum related to a technology that has substantial, ongoing, and potentially irreversible economic consequences would surely fall within Congress’s expansive mandate.

One other preliminary question -- is there any historical support for Congress exercising such power? Yes, quite a bit. Throughout history both Democrats and Republicans have considered and introduced legislation advocating for a national referendum on important policy questions. Nearly a century ago, Democrats weighed asking the American people if they supported the nation joining the League of Nations. In 1964, Rep. Charles Gubser, a Republican from California, sponsored a resolution to hold an annual nationwide opinion poll on key policy questions.

Even high-ranking officials have recognized the viability and value of a national referendum. Case in point, in 1980, House Majority Leader Richard Gephardt proposed a modified version of Rep. Gubser's idea--calling for a biannual poll on three designated issues. It’s also worth noting that many Americans are accustomed to voting on initiatives and referendums; a majority of states afford the public some form of direct democracy.

Finally, the most important question: why is an AI referendum necessary? First, the nationwide impact of AI on nearly every aspect of our day-to-day lives--from education to health care, the economy to transportation--makes this a question too big to leave up to a handful of tech billionaires and career politicians. Though the referendum would be nonbinding on Congress, the results would give voters a chance to see if their representatives actually listen to their constituents.

Second, placing a series of questions pertaining to what values and goals should inform AI regulation would spur more concrete discussions on the topic. For instance, we may never have precise estimates of which professions will be displaced by AI and when, but surely we can and should try harder to provide the public with such information so that they can see if the supposed benefits of AI really outweigh the costs.

Third, this approach would prevent Congress from getting ahead of itself (and the public) by enacting legislation that not only diverges from the will of the public but also has long-term and irreversible unintended consequences. Big regulatory undertakings are akin to aircraft carriers--hard to steer in a new direction.

Rushing to regulate AI is not only unwise from a policy point of view; it’s also profoundly democratic. Let’s give the people a chance to directly inform how Congress governs what may be the most consequential technological advance of our time. #LetUsDecideAI