Election Forensics

RJ O Taduran

One of the crucial differences between a police detective and a forensic scientist is the approach—given the available pieces of evidence, the former would try to identify who is/are the perpetrator(s), while the latter would try to figure out how an incident of interest happened. So yeah, that classic rock anthem “Who are You” by The Who may not be an appropriate theme at all for a popular television show about forensic scientists.

The 2022 Philippine election season has made more prominent the field of forensics, mainly because of debates about crowd sizes in campaign rallies. To determine if photos of large gatherings are real, edited, or fake, enthusiastic people have performed digital image analysis with image editing software and/or websites, where one can upload photos and do different layers of inspection for free. Some of the ‘forensic examinations’ of the questioned images are really impressive, as they demonstrated proficiency in art, photography, even in physics.

There have been many disputes about crowd estimates as well. To find out crowd density and uncover any misrepresentation or overestimation, passionate folks have conducted their very own scrutiny by using Google maps and other readily available websites to prove or disprove assessment provided by the police or the political party involved, while relying on their aptitude on geography, proxemics, and statistics. Forensics, as it should be, is interdisciplinary in nature—the more fields in the mix, the
better and more reliable the analysis.

How can one determine if there are illegal interferences or election manipulation? Experts in election forensics prefer to employ various statistical methodologies developed in the field of forensic accounting. Poll data are numerical, and the fundamental principle is that votes should follow naturally occurring patterns produced by a natural process, as in the case of a free and honest election. The main goal is to
detect deviations from what is expected to happen, because these would indicate alteration or manipulation of the voting results. Electoral fraud, just like any other crime, leaves distinctive traces.

In this time of automated elections, votes are considered big data that upsurge in terms of volume, variety, and velocity, as each vote counting machine or VCM can accommodate 800-1000 voters. One of the more preferred quantitative methods in election forensics is called Benford’s Law. It is a probability distribution for the likelihood of the first digit in a set of numbers, but it can also be used to predict the distribution of second digits, third digits, digit combinations, and so on. Benford is widely used in auditing to identify “doctored” numbers.

Another approach detects incremental and extreme fraud from the concurrent statistics of vote and turnout numbers. According to a study led by complex systems scientist Peter Klimek: “Incremental fraud means that, with a given rate, ballots for one party are added to the urn and/or votes for other parties are taken away. Extreme fraud corresponds to reporting a complete turnout and almost all votes for a single party.” Incremental fraud could be what we call dagdag-bawas in Philippine elections.

It is possible to hack an election because there are weaknesses in any automated system. In this time of electronic voting technology, one would need computer experts to confirm or refute any hacking events. While the Commission on Elections (Comelec) has assured the public that the data leak in the Smartmatic system was not related to the election, some are still afraid of any possible electoral fraud via cyberhacking come Election Day. Election officials should consider forensic examination when there
appears to be failure in terms of accuracy, availability, secrecy, and anonymity, as stated in a technical paper by computer scientist Matt Bishop of University of California, Davis and colleagues.

“Attackers could infiltrate election-management systems,” warned University of Michigan’s J. Alex Halderman in his interview with Scientific American. Malicious computer programs can be implanted in voting machines, which could lead to incremental or extreme electoral fraud. But just like what I previously said, electoral fraud, like any other crime, leaves distinguishing traces. Computer programs follow certain commands, and those commands would show signs of patterns that are not
natural, patterns that are not human-like.

This is why election forensics would need an expert in human behavior as well. Amidst all the statistics and computer science, the human behavior expert can answer the questions related to the patterns revealed by the quantitative methodologies. Can a group of people in a certain environment vote this way? Humans are bad at being random, yet they do not follow rigid patterns (of behavior) like a computer program would do. Humans continuously learn, and learning as course of adaptation has
increased during human evolution. This makes our species not too random nor too predictable, but just natural.

Clean and fair elections are the foundation of every democratic society. Election Day is a day when all of the citizens are truly equal. I hope there will be no need for election forensics on May 9, but as the age-old Pinoy joke would go: “Walang natatalo sa eleksyon, ang meron lang yung nananalo at yung nadadaya.”

Dr. Richard Jonathan O. Taduran is a forensic scientist, with specialization in biological and forensic anthropology.