Why AI Debates Should be Illegal

The image at the top of this article is in fact real. Though you might mistake it for a comedic skit, John Reid — the GOP candidate for Lieutenant Governor — really did hold a live debate against an AI-generated version of his opponent this past Tuesday. Reid resorted to this creative stunt after his opposition, State Senator Ghazala Hashmi, had turned down all requests for a debate.

Beyond the hilarity of the spectacle, an AI imitation of a potential political candidate prompts serious questions about how the technology should be integrated into our political systems. Civically involved citizens should be against this latest involvement of AI, as the creation of a properly sourced AI bot is subjective, the debate form can be manipulated, and because of the ethical problems in normalizing supposedly accurate AI depictions of political figures.

There is firstly the problem of what an AI bot is fed to create an accurate portrayal of another’s political position. Reid’s campaign claimed that it “largely gathered from previous interviews and her campaign website,” in order to make its depiction of Hashmi. Yet exactly what was pulled is not known. The slippery slope is apparent — if politicians can cherry-pick interviews, quotes, and data to feed an algorithm, without ever having to fully explain the process in doing so, politicians can easily misconstrue their opponents.

For example, there was a moment in the debate where the AI was questioned about the violent texts from prospective Democrat Attorney General Jay Jones, and after being asked for a clarifying response about whether he should drop out of the race, the bot offered no response. This is in a way not a lie — Democrats across the ballot have condemned Jay Jones’ rhetoric without saying he should actually drop out of the race. Yet it is nevertheless off-putting for a politician to have a representation of them make such a damning non-comment. It is highly doubtful that Hashmi would have stayed silent in an actual debate.

In addition to potentially creating non-genuine representations, AI debates allow the creator to manipulate the format to their advantage. For example, Reid’s debate consisted of an AI moderator asking a question to the AI Hashmi, Hashmi offering a relatively short response, and Reid having the opportunity to rebut and offer his position, before finally moving on to the next question.

The whole setup didn’t allow for any responses by the Democratic candidate. For viewers trying to form an opinion, this debate format would probably leave them convinced that Reid had all the right answers. In actual debate, where candidates have to wrestle over the feasibility of each other’s positions, viewers are informed in a way that isn’t one dimensional. As our society careens toward polarization, the art of real time debate will be lost, only pushing voters and politicians into echo chambers.

The most concerning consequence of expanded AI debates is an ethical one. While Reid made it clear that his AI debate was not the actual Hashmi, and cleverly used it to attack her no-show for an actual debate, we will eventually live in a world where it’s not so easy to tell that the AI model is not the real Hashmi. Even if stunts like these are pulled jokingly, it is easy to imagine a world where short-form content is flooded with believable clips from debates that never happened.

These could come from politicians themselves, their supporters, or foreign governments interested in influencing US elections. The moral problem of purposefully making AI in the image and likeness of others — without their consent or knowledge — could lead to a political landscape where it’s almost impossible to tell what the real positions of our politicians are. In this case, Reid communicated well about how his debate did not have this intent, but the ethical issues the format poses still remain.

It is a sad reflection of our political system that politicians no longer feel the need to argue for their positions. The art of a good debate should be a cornerstone of the US political system and should help form the opinions of the electorate. Using AI to imitate such debates should be illegal, and politicians should be proactively thinking of ways to prevent its proliferation. Mandating in-person debates would also eliminate the need for stunts such as these and help to preserve the fabric of American democracy as we enter an era of ubiquitous AI.