Combating Electoral Fraud with Artificial Intelligence
Photo Source: Wikimedia Commons
Voting represents the essence of democracy. The freedom to privately cast a vote gives every citizen a voice in the governing of their country, state, or town. Democracy needs voting to work and equally depends on the truthful execution of vote counting after an election to survive. Confidence in the voting process leads to peaceful transitions of leadership, and conversely, suspicion of electoral fraud leads to unrest and, in some cases, civil war. The highly controversial presidential election in 1876 between Samuel Tilden and Rutherford B. Hayes saw Tilden win the popular vote overwhelmingly but lose by one electoral vote in a deal called the Compromise of 1877. The Compromise gave Hayes twenty disputed electoral votes in exchange for the final withdrawal of all Union troops from the South. More recently, the Nigerian presidential election of 2007 involved violence, vote rigging, and ballot stealing. An article titled “Nigerian election pushed back a week,” by CNN wire staff described the election as “marred by riots, bombings, and assassinations.” Suspicion of election fraud tainted a number of elections around the world from Russia to Venezuela over the past twenty years.
With voter fraud and election rigging threatening to undermine democracy at every election, researchers, mathematicians, and data scientists continuously work to develop ways to detect the occurrence of electoral fraud using mathematical models and, increasingly, artificial intelligence. A research paper by the New York University professor of politics, Arturas Rozenas, titled “Detecting Election Fraud from Irregularities in Vote-Share Distributions,” uses statistics to identify false reporting of election results. The technique uses historical election results from both known honest and fraudulent elections to identify cases where results do not match expectations. (Political Analysis, 2017) Ines Levin of the Department of Political Science at UC Irvine points out two types of statistically identifiable electoral frauds—vote stealing and ballot box stuffing. She shows how voter turnout follows a predictable pattern, and when vote stealing occurs the voter turnout appears much lower than expected with deflated vote shares for one party or candidate. On the other hand, with ballot box stuffing, voter turnout appears unusually high with vote shares favoring one party or candidate. Employing artificial intelligence, she used a mixture of data from fair and fraudulent elections to train her computer to detect voter fraud and then analyzed the Venezuelan presidential election results from April 2013. The artificial intelligence detected both vote stealing and ballot box stuffing in the Capital District. In an article titled “An Informed Forensics Approach to Detecting Vote Irregularities,” the authors admit that limitations of human bias and a lack of accurate data make election fraud exceptionally difficult to detect. To get around these limitations, they developed a machine learning technique called BART that uses results from several hundred elections from around the word to find patterns suggesting voter fraud based on the theory that specific numbers in election returns occur more frequently in honest election results than in manipulated elections.
Free and open elections form the basis of democracy, but electoral fraud continuously threatens to undermine the people's trust in the results. Moreover, with a divided populous, elections need to be conducted with extreme care to avoid any threat of voter fraud. A loss in confidence in election results can lead to civil unrest and in extreme cases violence and even civil war. Contested elections in many countries from Zambia to Hungary have incited violence and politically motivated arrests. Such threats to democracy make honest elections all the more crucial to a well-functioning and peaceful society. Detecting electoral fraud remains difficult, but researchers continue to develop new tools using mathematics and artificial intelligence to identify unusual patterns in election participation and returns that can suggest cases of vote manipulation. The ability of well designed artificial intelligence to take most bias out of election monitoring makes them a great new tool to safeguard our electoral process, but many researchers still admit that even the new machine learning tools remain insufficient for detecting all cases of voter fraud. Additionally, to maintain the peoples’ confidence in the electoral system, researchers must be careful not to make false positive claims of vote manipulation. Diligent election monitors combined with new artificial intelligence must work together to safeguard our democracy and society.
Dr. Smith’s career in scientific and information research spans the areas of bioinformatics, artificial intelligence, toxicology, and chemistry. He has published a number of peer-reviewed scientific papers. He has worked over the past seventeen years developing advanced analytics, machine learning, and knowledge management tools to enable research and support high level decision making. Tim completed his Ph.D. in Toxicology at Cornell University and a Bachelor of Science in chemistry from the University of Washington.
You can buy his book on Amazon in paperback and in kindle format here.