AI and US Elections

As the founder and CEO of a data and AI consultancy, I wonder about both the positives and the negatives of AI.  Recently, I have been thinking about the role AI might play in the upcoming US elections.

Deepfakes.  AI’s ability to create very realistic pictures and videos is an area of concern for elections.  This two-year-old video of a Morgan Freeman deepfake illustrates the risk.  Could a deepfake video of a candidate saying or doing something inappropriate go viral in the days immediately preceding an election?  Sure.  However, I don’t see that as a major concern.  I think that any simulation of a candidate saying or doing something inappropriate would come under serious scrutiny, especially if the simulation went viral.  Between the campaigns and the media, I think a deepfake would be quickly debunked.

Microtargeting.  In a close national election, a very small percentage of the total voter pool, swing voters in battleground states, could decide the winner.  AI allows for very granular targeting of advertisements and messaging.  This combination of a relatively small number of voters and AI-based microtargeting could be troublesome.  The campaigns could use AI to inexpensively create a wide variety of political message ads and the use AI to feed the right ads to the right swing voters.  A swing voter in a farming community might see and ad that emphasizes how the candidate supports the mandated use of ethanol (made from corn) in motor fuel.  A swing voter in an urban setting might see an ad from the same candidate decrying the government’s use of tax dollars to subsidize business.  Both voters are left believing that the candidate supports their position even though ethanol mandates are a form of government subsidy.

Social media.   A recent study detected 1,305 unique bot-based accounts operating on X and Reddit during a recent Republican primary debate.  AI, coupled with social media, is being used to increase divisiveness.  AI-based sentiment analysis can find the social media interactions where “us vs them” is prevalent.  Generative AI can automatically write social media posts that help reinforce echo chambers and add to overly emotional debates.  AI makes social media bots more powerful and sophisticated.

Chatbot disinformation about voting.  Intentionally erroneous information can be generated by chatbots disguised as sites intended to answer questions about where and when to vote.  The pandemic-fueled changes in voting rules coupled with a very decentralized approach to voting in the United States leaves many voters with questions.  Where do I vote?  When can I vote?  Can I use mail in ballots?  Chatbots that seem to provide voters with voting information could mislead voters to suppress the vote in certain geographical areas.  In the 2022 mid-term elections, people in Iowa were told they could vote over the phone.  Advances in voice synthesis will make such fraudulent disinformation schemes even more believable.

Despite the possibility of AI-generated manipulation of the of the upcoming US elections, I remain positive.  The Taiwanese elections of Jan 13 give me reason to be optimistic.  Despite widespread attempts by the People’s Republic of China to influence the Taiwanese vote, the candidate least favored by the PRC won the election.

It seems like voters are still a step ahead of the AI.

 

Denis McFarlane, CEO

Are you ready to get more value out of your data?