Mind Matters News and Analysis on Natural and Artificial Intelligence
Helmeted police officers photographed from behind during a protest

Can AI Predict and Prevent Political Unrest?

The 1996 Democratic Convention tried neural networks but discovered a hidden flaw

When the Democratic Party held the national convention that nominated Bill Clinton for president in Chicago in 1996, it sought to prevent a repeat of the rioting that accompanied the Convention of 1968. To help with that, the Internal Affairs Department of the Chicago Police turned to neural networks for help.

Neural network software programs recognize patterns. The pattern the police department wanted to identify was behavior on the part of police officers that would likely intensify demonstrators’ violence.

What went wrong in 1968? First, it was a volatile year. A few months before the convention, civil rights leader Martin Luther King and presidential hopeful Bobby Kennedy were both assassinated. In the aftermath, Kennedy’s assassination opened the way for the nomination of the sitting vice president Hubert Humphrey. Humphrey ultimately lost the election to Republican Richard Nixon.

Meanwhile, protests against the Vietnam War were everywhere, accompanied by colorful street drama. Anarchist Yippies were amusing but ultimately dangerous. They spread their strange brand of chaos by giving away bologna sandwiches outside of thousand dollars- per-plate political fundraisers. From the spectator balcony of the New York Stock Exchange, they mockingly showered traders with fluttering dollar bills and chuckled as they scrambled for the free money. They spread rumors that smoking dried banana peels could get you high. The pop singer Donovan popularized the rumor with his hit Mellow Yellow, whose lyrics included the line “Electrical banana is gonna be a sudden craze.”

The Yippies staged noisy protests against the Vietnam War outside the 1968 Democratic National Convention. What motivated them? When the motorcycle outlaw played by Marlon Brando was asked in the 1953 outlaw gang classic, The Wild One what was he rebelling against, he replied, “Whaddaya got?” That summed up the Yippies.

But this time it was not just a film. The protesters butted heads with political strongman and legendary Chicago mayor Richard J. Daley. Daley thought the protestors fouled his city and he released his police department like a pack of attack dogs. On the streets outside of the Chicago’s International Amphitheatre where the Democratic convention was being held, violent clashes between the protestors and the Chicago police resulted. A presidential commission later called the conflict a “police riot.”

According to one account,

After four days and nights of violence, 668 people had been arrested, 425 demonstrators were treated at temporary medical facilities, 200 were treated on the spot, 400 given first aid for tear gas exposure and 110 went to hospital. A total of 192 police officers were injured.

Images of police firing teargas and beating demonstrators with their nightsticks played on network television news.

David Taylor and Sam Morris, “The whole world is watching” at The Guardian

The Chicago Seven trials following the convention featured Yippie leaders like Abbie Hoffman and Jerry Rubin, charged with conspiracy and inciting to riot. In the true spirit of anarchy. Abbie Hoffman later wrote Steal This Book (1971), a handbook of the Yippie culture. Tom Hayden, former husband of actress Jane Fonda, was also a member of this not-so-magnificent Chicago Seven.

Twenty-four years later, in 1996, another Democratic national convention was scheduled for Chicago. Richard J. Daley was no longer mayor, but his son Richard M. Daley was. What could be done to prevent the recurrence of a police riot?

The city government turned to a neural network for help. It was tasked with identifying bad cops who might become enraged and start bonking heads with billy clubs.

The neural network looked at 12,500 police officers. They were compared to about 200 officers who were previously dismissed or had resigned while under investigation. The charges against the troubled cops in the database ranged from insubordination to criminal misconduct. After the neural network was trained with historical data, a total of 91 of the 12,500 currently serving police officers were identified as a potential risk by the neural network. Those so identified were asked to enroll in a counseling program.

But the union representing the police officers challenged the machine’s obscure decree in court. Why these officers in particular? What had they done?

The police union’s objection pinpoints a weakness of neural networks even to this day. The neural network is basically a black box. When the computer programmer was asked why a cop was classified as a bad cop, the only explanation was “Because I trained the neural network to detect bad cops and the neural network did what I trained it to do.” That was not considered a good enough legal reason. Specifics must be cited, such as “The officer regularly attends dog fights” or “The police officer we fingered as dangerous regularly beats her husband.”

This conflict arises because a generic neural network lacks an explanation facility which assigns reasons as to why a conclusion is reached. Although there is some research into constructing explanation facilities for neural networks, the layered perceptron and the convolutional neural network used in deep learning remain fundamentally black boxes to this day.

In any event, the fuss around the 1996 Democratic National Convention turned out to be much ado about nothing. Abbie Hoffman had committed suicide in 1989. His Yippie co-founder Jerry Rubin had died in a traffic accident in 1994. Tom Hayden, after divorcing Jane Fonda, served in the California State Senate. The Vietnam War was long over and there seemed to be nothing much to protest except Bill Clinton’s extramarital affairs. So insofar as the 1996 Democratic National Convention was concerned, all went smoothly even though the neural network results were ignored.


Also by Robert J. Marks: Why it’s so hard to reform peer review. Robert J. Marks: Reformers are battling numerical laws that govern how incentives work. Know your enemy!

and

Study shows that eating raisins causes plantar warts. Sure. Because, if you torture a Big Data enough, it will confess to anything

Featured image: Riot police/Gerard Bottino, Adobe Stock


Robert J. Marks II

Director, Senior Fellow, Walter Bradley Center for Natural & Artificial Intelligence
Besides serving as Director, Robert J. Marks Ph.D. hosts the Mind Matters podcast for the Bradley Center. He is Distinguished Professor of Electrical and Computer Engineering at Baylor University. Marks is a Fellow of both the Institute of Electrical and Electronic Engineers (IEEE) and the Optical Society of America. He was Charter President of the IEEE Neural Networks Council and served as Editor-in-Chief of the IEEE Transactions on Neural Networks. He is coauthor of the books Neural Smithing: Supervised Learning in Feedforward Artificial Neural Networks (MIT Press) and Introduction to Evolutionary Informatics (World Scientific). For more information, see Dr. Marks’s expanded bio.

Can AI Predict and Prevent Political Unrest?