Fears of deepfakes derailing democracy are looming. Incidents such as the Joe Biden robocall debacle and various examples of deepfake videos depicting politicians have voters and lawmakers on edge, resulting in an explosion of legislative proposals to regulate AI-generated content.
In response to these notable episodes, the Senate recently “advanced bills” that would limit the use of deepfakes in election advertisements. These bills provide remedies by “barring deceptive videos and audio” of political candidates, labeling requirements and developing guidelines to mitigate the risk of AI.
The problem is that the concerns over deepfakes interfering with elections are overblown. If anything, voters are a greater danger to democracy than deepfakes.
Since voters find deepfakes that reinforce their beliefs more convincing, regulating deepfakes will limit political speech, and voters are notoriously uninformed.
Despite the lifelike realism of deepfake media, people are surprisingly good at detecting them. One study found that when political deepfakes were shown, 53 percent of participants correctly identified them as fake. Low-tech forms of deceptive media, false headlines, were marginally less persuasive than deepfakes. Another study by MIT discovered that 70 percent of the participants identified a deepfake of Vladimir Putin. In contrast, the detection model failed to determine it was a deepfake.
Partisanship plays a role in voters finding misinformation believable. Studies have also established that people are likely to label a video as fake if it makes their political party “look bad.” This is true of Republicans and Democrats.
Plus, media literacy has done little to curb false media, and “many people” have knowingly “engaged” with deepfakes. The danger weighs more heavily on weaponized false media reinforcing partisanship through confirmation bias.
These bills are most likely a response to the public’s concern over false depictions of political figures. Eighty percent of voters are concerned about political deepfakes. This public outcry for regulation might hamper the right to free expression. Favoring policies that limit free speech is anti-democratic because it has a chilling effect on discourse and limits the spread of ideas.
Regardless of the popularity of regulating misinformation, false speech is constitutionally protected. The Supreme Court decision U.S. v. Alvarez “expressed that content-based restriction is only valid if the state has a compelling interest.” Otherwise, the government risks engaging in viewpoint discrimination, violating the First Amendment. The laws barring AI depictions of politicians could apply to constitutionally protected satire. If you can’t make fun of elected leaders, you truly live in a dictatorship.
Even before AI-generated election ads, democracy faced the challenge of uninformed voters. The theory of Rational Ignorance seeks to explain why voters choose to remain uninformed.
This theory suggests the costs of being fully informed on political issues outweigh the direct benefits. We live in a society where only 34 percent of Americans can name the three branches of government. Sixty-three percent of voters don’t know who their representatives are. Thirty-seven percent are unaware of the party that controls the Senate. Never mind Americans reading political journals to stay informed; the number of people following the news is plummeting.
So what? What is the harm in letting people vote on emotion versus facts? By doing so, voters unwittingly impose bad policies on constituents.
Beyond narrow policies, voter ignorance can be a direct assault on democracy. Interest groups and opportunist politicians can “take advantage” of the public’s lack of political knowledge. Especially in times of crisis, out of trepidation, support anti-democratic policies. For example, politicians capitalizing on the border crisis as a pretext for widely supported voter ID requirements.
Although in-person voting fraud is rare, ID requirements could serve as a springboard for more restrictive voting laws. Partisanship and ill-informed voters are a toxic brew for democracy.
Making deepfake technology the scapegoat will only compound matters by restricting the marketplace of ideas. Deepfakes are a minuscule threat.
Careful research and civic participation can reduce the slight risk of interference by AI. It may reduce the tide of Americans gravitating toward social media posts reaffirming their beliefs.
