Pundits Vs. Machine: Predicting Controversies In The Presidential Race | KUOW News and Information

Pundits Vs. Machine: Predicting Controversies In The Presidential Race

Sep 19, 2016
Originally published on September 20, 2016 6:41 am

Predictions are for psychics — and in this very unpredictable political season they might do a better job than the pundits. But what about a computer? I set out to see how well it could predict which controversies around the candidates were likely to re-emerge over the course of a month. And two human pundits have agreed to compete against the machine.

Meet the Contestants

The Computer

The computer is run by Quid, a data analytics firm that uses proprietary software to search, visualize and analyze text. Since the computer can't speak, Dan Buczaczer, Quid's head of marketing, is going to speak for it and explain how it "thinks."

"Quid uses proprietary software to search, visualize, and then analyze massive amounts of text," Buczaczer says. "In this case, what we're talking about today, that massive amount of text happens to comes from news sources and blogs written about a particular topic, in this case the presidential election."

And Buczaczer is talking about some 300,000 U.S. blogs and publications amounting to nearly 7.5 million articles — everything that has been written about Hillary Clinton and Donald Trump since they announced they were running for president.

Quid's computers sifted through all that coverage to find every controversy that has plagued each candidate. "Two-thirds of them were Donald Trump, one-third Hillary Clinton," Buczaczer says. "So Trump is the winner in terms of overall number of controversies generated." Though Clinton had fewer controversies, they still generated as much coverage as Trump's.

The computer sifted through all them for patterns — like, which ones kept re-appearing.

"We kind of mapped it against both reoccurrence and importance — what sort of an impact did it have at its peak?" Buczaczer says. "In a lot of ways this was probably the heaviest part of computation around what we think is going to show up again and again for each candidate."

Buczaczer and his computer have done some forecasting about which controversies will get the most coverage for each candidate between Sept. 12 and Oct. 12. But I'm not going to tell you what they are — not yet.

The Humans

In the right corner: Jonah Goldberg, who writes for the conservative National Review.

"This is a very intimidating thing to do because obviously you're putting me up against Skynet," Goldberg says.

Skynet is the computer in the Terminator movies that ends civilization as we know it. While Quid was searching through millions of articles looking for patterns — Goldberg has a much more poetic way of predicting the future.

"It's always safer to bet making predictions that are in line with the character and personality of the people you're making predictions about," Goldberg says. "Like in Aesop's fable, it's really easy to predict that the scorpion is going to sting the frog — because that's the scorpion's nature."

In the left corner: Simon Maloy, a political writer at the liberal Salon.

Maloy agrees that the personalities of the candidates are part of his prediction process. But so is the media itself. It has to draw readers and viewers and create buzz, which is why the media likes controversies.

"I think media organizations understand that there's value in writing a story that people will fight about and people will talk about. I think that absolutely factors into it," Maloy says.

The Rules of the Game

Each contestant will predict which controversies will get the most coverage over a month-long period. They each gave a list of predictions for each candidate. The list is ranked in order of which controversy will get the most coverage.

On Oct. 12, Quid's computers will search the Web to find out which controversies got the most coverage. And we will announce the winner on air and on this blog. Each contestant will get their own NPR T-shirt. Then, we will do some deeper analysis of other areas where both humans and machines are used to make predictions.

The Predictions

Copyright 2016 NPR. To see more, visit http://www.npr.org/.

ROBERT SIEGEL, HOST:

Predictions are for psychics, and in this very unpredictable political season, they might do a better job than the pundits. But what about a computer?

(SOUNDBITE OF MUSIC)

SIEGEL: On today's All Tech Considered, NPR's Laura Sydell pits a computer against two political columnists to find out who or what can best predict the political scandals we'll be talking about in a month.

LAURA SYDELL, BYLINE: First let's meet the contestants. In this corner...

DAN BUCZACER: Dan Buczacer - I run marketing at Quid.

SYDELL: Quid is a data analytics firm. Buczacer is talking for the computer because it doesn't speak.

BUCZACER: Quid uses proprietary software to search, visualize and then analyze massive amounts of text. In this case, what we're talking about today, that massive amount of text happens to come from news sources and blogs written about a particular topic, in this case the presidential election.

SYDELL: And we're talking about some 300,000 U.S. blogs and publications amounting to nearly seven and a half million articles, everything that has been written about Hillary Clinton and Donald Trump since they announced they were running for president. Quid's computers sifted through it to find every controversy that has plagued each candidate.

BUCZACER: Two-thirds of them were Donald Trump, one-third Hillary Clinton. So Trump is the winner in terms of overall number of controversies generated.

SYDELL: Though Hillary Clinton had fewer controversies, they still generated as much coverage as Trump's. The computer then sifted through all of them for patterns, like which ones kept reappearing.

BUCZACER: We kind of mapped it against both reoccurrence and importance. What sort of an impact did it have at its peak? And so in a lot of ways this was the probably the heaviest part of computation around what we think is going to show up again and again for each candidate.

SYDELL: Buczacer and his computer have done some forecasting about which controversies are going to follow each candidate the most between September 12 and October 12. But I'm not going to tell you what they are - not yet. First let's meet our human contestants.

JONAH GOLDBERG: This is a very intimidating thing to do because I mean obviously you're putting me up against Skynet.

SYDELL: That's Jonah Goldberg of the conservative National Review, and Skynet is the computer in the "Terminator" movies that destroys civilization as we know it. While Quid was searching through millions of articles looking for patterns, Goldberg has a much more poetic way of predicting the future.

GOLDBERG: It's always safer to bet making predictions that are in line with the character and personality of the people you're making predictions about. Like in "Aesop's Fable," it's really easy to predict that the scorpion's going to sting the frog because that's the scorpion's nature.

SYDELL: And our final contestant on the left...

SIMON MALOY: My name is Simon Maloy, and I am a political writer for salon.com.

MALOY: Maloy agrees that the personalities of the candidates are a part of his process, but so is the media itself, which has to draw readers and viewers and create buzz, which is why the media does like controversies.

MALOY: I think media organizations understand that there's value in writing a story that people will fight about and people will talk about. I think that absolutely factors into it.

(SOUNDBITE OF MUSIC)

RALPH: And now it's time for computer pundit versus human pundit, where our contestants see who or what can pick the top controversies. Take it away, Laura.

SYDELL: Thank you, Ralph. Each contestant will predict which controversies will get the most coverage over a month-long period. They gave their predictions to NPR on September 12. So the game ends October 12. Let's start with the machine and its forecast about the coverage of Donald Trump.

BUCZACER: Building a wall along the Mexican border.

SYDELL: Yes, it's the wall.

(SOUNDBITE OF BELL)

SYDELL: And now onto our first human over here on the right, Jonah Goldberg. Number one...

GOLDBERG: Is a lot of discussion about his tax returns.

(SOUNDBITE OF BELL)

SYDELL: And our last contestant's forecast for Trump, Simon Maloy.

MALOY: I think Trump will at some point say something sexist about Hillary Clinton.

(SOUNDBITE OF BELL)

SYDELL: And now round two - our contestants will make their top predictions about Hillary Clinton. Let's start with the computer.

BUCZACER: With Hillary, I will be shocked if anybody that you're talking to has a different number one. And if they do, I predict they will lose. At number one for Hillary is her email scandal of course.

(SOUNDBITE OF BELL)

SYDELL: And now human number one, Jonah Goldberg.

GOLDBERG: Debate mishap - I think it would be inappropriately laughing at something she shouldn't laugh at, or she does it so much it kind of goes viral the way Al Gore's sighs in 2000 during the debates became a thing.

(SOUNDBITE OF BELL)

SYDELL: And Simon Maloy with our final one for Hillary Clinton.

MALOY: There'll be a ton of coverage the moments that Clinton makes her first public comments on her health status. Whatever she says, I think it's going to be pored over and picked apart endlessly compare with past statements.

(SOUNDBITE OF BELL)

SYDELL: And folks, that concludes our contest - to humans versus a machine. They each made many more predictions in order of what gets the most coverage, and you can find those on our website.

(SOUNDBITE OF MUSIC)

SYDELL: Candidates will be judged by the machine which will measure how much coverage each issue got. Don't worry. It's not programmed to cheat. And Ralph, tell us what each contestant wins.

RALPH: Guests on computer pundit versus human pundit all receive an NPR T-shirt and a badge to get them into the NPR headquarters for three quarters of an hour.

SYDELL: And in the week of October 12, NPR will let you know the winner. For political pundits, this could be a turning point. If they haven't given up making predictions yet, this could be the end. Laura Sydell, NPR News. Transcript provided by NPR, Copyright NPR.