AI, misinformation, foreign actors … what could go wrong?
Since 1796, Americans have had presidential election campaigns every four years. But in recent years, technology has changed everything.
Campaigns started with anti-Federalist articles in daily newspapers. The 19th century ushered in smear campaigns. By the 20th century, television, the newest and biggest technology at the time, changed campaigning forever. Americans watched intently from their living rooms as candidates debated with one another on pivotal issues.
In the new age, politicians rely on the Internet and artificial intelligence (AI) to spread their message and reach voters through their websites and social media sites. AI combines a large amount of data with algorithms to simulate human intelligence in machines and other software.
The Good
AI can help partisans better listen and serve the public by analyzing their online behavior, data consumption, relationships, and social media patterns to create unique behavioral profiles. These profiles can then help their campaigns run personalized microtargeted advertising campaigns to help voters make an informed decision.
Social media platforms use machine learning to analyze data and identify patterns so they can target an ideal group without any human intervention. The computer will scan the data until it finds a repetitive structure and creates an output. For example, Facebook allows candidates to target small groups and individuals for microtargeted advertising campaigns through its “custom audience” feature. Candidates can also use a “look-alike” tool to help find people who will match their ideal voter.
Technology has also allowed diplomats to rely more on insight rather than instinct by using big data—the technology behind analyzing large data sets with the use of traditional data processing application software as well as social media and AI systems. In fact, The Economist did just that by collecting polling, economic, demographic and historical data into a machine learning code to predict statistics during this year’s election.
Another prediction model reveals that Donald Trump has a 91 percent chance of winning the election. Helmut Norpoth, a professor of political science at Stony Brook University and one of the few to correctly predict the outcome of the 2016 U.S. Presidential election, states that Trump will get 362 electoral votes and Biden only 176.
According to Norpoth, his Primary Model is based on statistics dating back to more than a century and relies on the presidential primary, where voters participate in choosing the candidate for their party’s nomination in a general election. Since the beginning of primary elections, candidates with the better primary vote have tended to win the general election, which in this case was Trump. Norpoth states that the forecast, posted on March 2, is unconditional and final.
Caucuses are also using a machine learning technique called scoring to rank voters from 1 to 100 on how likely they are to do something or hold a certain opinion. They then use this data to persuade undecided voters and inform them of their beliefs.
AI and machine learning tools are also helping Americans discover who they should vote for based on their answers to questions on topics such as immigration, the environment, foreign policy, and other important issues. One of these tools is iSideWith.com. The tool has already attracted over 50 million users to complete the online quiz and is making elections more democratic by matching voters to candidates based on their views.
Such tools also help legislators make decisions about different issues around a city. For example, policymakers in Austin, Tex. purchased data from iSideWith to help the city council understand the issue of ride-sharing apps and how voters felt about them rather than depending on feedback at public meetings.
The Bad
As with most technology, the use of AI during elections has led some to question whether the technology is truly advantageous.
Some evidence shows that AI can manipulate citizens by spreading propaganda and fake news on social media. Since bots or autonomous accounts are programmed to spread one-sided political messages, it can lead to bias and influence the general public by discouraging them to vote for the opposing side or causing them to stop voting altogether. Not to forget, these bots have developed a sort of persona with fake names, bios and photos generated by AI, so they can go easily undetected by the network and users reading their messages.
The University of Washington revealed that during the 2016 U.S. presidential election, automated social media bots were using Twitter to generate tweets using pro-Trump hashtags and promoting anti-Hillary Clinton content. His party also used machine learning to gain data from over 87 million Facebook users and display personalized content to create a pro-Trump bias.
During the same election, pro-Clinton campaigners spread automated content to generate a quarter of the traffic on Twitter and throw off the support from the Trump campaign. An extensive advertising campaign by pro-Clinton campaigners targeted voters based on psychology by using big data and machine learning to influence people’s emotions. Different voters received different messages based on AI’s prediction of their susceptibility to different arguments.
The use of AI in both cases led to false political messages and a general loss of faith in the political system.
Viral fake news, photo scams and other misleading information can also negatively influence voting. Software such as DeepFaceLab and Faceswap allows anyone to create deep fakes, synthetic media in which a person is replaced with someone else to misinform and manipulate citizens. Since it’s easy to create fictitious content, there is a lot of it on the Internet that has led to controversies and political divisions.Deep fakes are made using a type of machine learning architecture known as a generative adversarial network, or GAN, and graphical rendering. The simulation takes in a large amount of data, including images and videos, to learn the key characteristics of the person and generate a new video with fake content that looks exactly the same.
Deep fakes can often be indistinguishable from real videos. In fact, according to the Brookings Institution, deep fakes can distort the democratic discourse and lead to public distrust in public institutions. Almost two-thirds of the U.S. population say that fake content creates a great deal of confusion about political realities to the point where many states have banned them during elections, including California, according to Pew Research Center.
A true democracy relies on fair elections where the citizens can vote without intimidation or manipulation, but with the rise of AI, democracy could be in jeopardy.
What’s Next in Technology
Technology forces us to adapt, often in ways that we don’t expect. Working alongside big companies and political bodies, society as a whole can further expand the benefits of technology to include ways to restrict misinformation, propaganda, and fake media as well as increase regulation and improve democracy.
Artificial personas are the future of propaganda. AI will soon be able to pose as individuals while writing intelligent letters and comments on political issues through various mediums. Ultimately, these personas will be able to debate and lead debates on the Internet.
The important takeaway from history—and especially the 2016 U.S. presidential election—is that it is cheap and easy to misinform the general public and future technology will make this even easier. Thus, politics could be full of intimidation and manipulation, potentially making the process undemocratic.
Voters themselves must demand facts and truth in online discourse. Otherwise, regulation may be key.
A proposed California law would require bots to identify themselves, which many states will soon follow suit. Several bills in Congress reflect this sentiment, like the bipartisan Designing Accounting Safeguards to Help Broaden Oversight and Regulations on Data Act and the Banning Microtargeted Political Ads Act. These bills are expected to be addressed in 2021.
Many people have also suggested replacing politicians or making policy decisions using AI alone. There have been many cases of AI bots running for office, including in Russia with Alice, in Tokyo with Michihito Matsuda, and in New Zealand with SAM. Though none of these bots won their elections, they still shared their thoughts and values with society and garnered a fair number of votes.
Since AI can make policies by gathering data from all Americans, citizens will be responsible for all legislative decisions, with digital agents voting on their behalf. This could boost citizen engagement and increase democracy.
Technology is changing the way people think and participate in a democratic society and it has the potential to, on some level, deepen democratic practices.