Free Republic
Browse · Search
GOP Club
Topics · Post Article

Skip to comments.

Trump's victory: the night a machine predicted humans better than the humans
Campaign ^ | November 9, 2016 | Lisa De Bonis

Posted on 11/22/2016 1:45:30 AM PST by 2ndDivisionVet

President Donald Trump.

It's fair to say pollsters, political commentators and experts on both sides of the Atlantic didn't believe they would ever see that. And neither did I.

If you listened to us all on the airwaves it was clear that most believed common sense would prevail and prevent someone so polarising and unpredictable being elected to the most powerful office in the world.

But there was one expert in a TV studio that, if it had had a voice, would have been saying something like "I told you so".

EagleAi was designed and built by Havas for ITV News. A learning artificial intelligence tool, it was programmed to interpret and understand on a scale never seen before on British television.

Our team at Havas Cognitive that created EagleAi fed it with billions of data points dating back to the moment Hillary Clinton and Donald Trump were selected at their respective party conventions. And then we spent weeks checking and double checking the data sets and the algorithms, because we just couldn't believe our eyes. You see, EagleAi kept predicting a Trump victory.

It was taught to understand sentiment, tone, emotion and even intention and use all of that to predict voter behaviour. In fact, it got so good at it, it could spot patterns and connections just like a human brain.

I spent election night chaperoning EagleAi on ITV News, and I was in awe of its ability from the start. This cognitive technology allows us to understand like a human being but operate at a volume and speed that is not humanly possible. It's like harnessing the power of 2,000 data researchers at the touch of a button.

After feeding on the content of countless speeches, three presidential debates, 15 million articles, hundreds of political feeds, tweets, social posts and millions more sources, this extraordinary piece of AI – the first of its kind to appear live on television – was telling its programmers that it believed Donald Trump in the White House was not just a possibility, but a likely outcome of this presidential election.

ITV News anchor Tom Bradby was incredulous when EagleAi produced the insight that Donald Trump's personality index score was in fact more "agreeable" than Hillary Clinton's.

EagleAi used personality indexing to anaylse elements of the candidates' character and their messages, from their debate performances, to their social channels and the content of their speeches.

The index suggests that a person is "agreeable" when they are perceived as warm and considerate. So did EagleAi spot something that the media didn't? That Trump's repeated use of the messages that he "loves his country" and wants to "make it great again" gave voters that exact impression – of someone who understood them and cared about them? No matter his repeated and increasingly bitter attacks on his opponents and minorities of all kinds.

What's more, EagleAi spotted something about Hillary Clinton too – that she may be prepared, diligent and on top of the detail, but when it came to warmth and likability it all felt too rehearsed and cold for the former Secretary of State to win over voters. And so her messages just didn't cut through like they should have done. Even though her credentials on civic issues like gender equality, abortion rights and civil liberties should stack up, she just wasn't offering the warmth that voters craved.

When EagleAi was asked to explore the election issues it had uncovered, it could offer so much more than polling data. In this election it was clear to us and the AI that volume was the number one priority. Say it often enough and people will hear.

And that's where Trump may have just won this election: the power of saying it over and over again, talking directly to the voters on social media, and then never letting up. Did the Clinton camp simply fail to respond to the reality of social media as the new powerhouse in politics, in brands and in business?

And that is where we believe cognitive is only at the beginning of its evolution. It is essential to the future of our industry and the understanding of brand penetration and ultimately buying behaviour.

If EagleAi could see what so many human beings couldn't when it came to the ragingly unpredictable nature of voting in an historic election, what might it be able to do for businesses all over the world just beginning to understand its power?

As EagleAi delivered the fatal blow to the Clinton campaign – showing a map of the US and support by state that had whole swathes of America bedecked in red, it displayed a laser-like clarity that the human beings hadn't.

And maybe that's the beauty of AI. It doesn't have a vested interest. EagleAi had no in-built bias on election night. When you focus on the zeros and ones that are the building blocks of code, there's no room for preconception or prejudice. Just the facts. On an awe-inspiring scale.


TOPICS: Campaign News; Polls
KEYWORDS: 2016polls; ai; analytics; eagleai; hillary; trump; trump2016; warmth
Navigation: use the links below to view more comments.
first previous 1-2021-25 last
To: spintreebob
Nice analysis. The author's bias was also shown in "And her messages just didn't cut through like they should have done." The objective is "could have done".

In an election this close with 4 major states at around 1%, how did it predict state-by-state? I dare say few if any got every state right, especially with WI going Trump and NH Clinton.

21 posted on 11/22/2016 5:33:33 AM PST by The Truth Will Make You Free
[ Post Reply | Private Reply | To 20 | View Replies]

To: The Truth Will Make You Free

About Halloween, I and many others predicted IA, FL, OH would go Trump but close. We thought NC, CO, NV, MN would go Clinton but close. ME would split.

We thought PA, MI, WI, NH were too close to call... at least from our arm chairs a thousand miles away.

We were wrong about OH. It wasn’t close.
We were wrong about NC. Trump won it.

Nate Silver’s 538 was wrong in FL, NC, ME, PA, OH, MI, WI, MN and probably more. His dark Blue should have been light blue in some states that Clinton won. His light red should have been dark red in some states that Trump won, eg OH.


22 posted on 11/22/2016 8:48:11 AM PST by spintreebob
[ Post Reply | Private Reply | To 21 | View Replies]

To: CARTOUCHE

That’s exactly who I meant. Thank you!


23 posted on 11/22/2016 9:05:13 AM PST by Fantasywriter (Any attempt to do forensic work using Internet artifacts is fraught with pitfalls. JoeProbono)
[ Post Reply | Private Reply | To 19 | View Replies]

To: spintreebob
Ironically, the one who seemed to get it right was garbage mouth Mike Moore who said the rust belt were going to vote Trump. Go figure. Even a broken clock is right twice a day.

Whether people believe in such things or not, there were also many reputable Christian prophetic voices that foretold of a Trump win. They were also correct, but didn't at all have insight on specific states.

24 posted on 11/22/2016 10:26:33 AM PST by The Truth Will Make You Free
[ Post Reply | Private Reply | To 22 | View Replies]

To: The Truth Will Make You Free

Scott Adams was certainly the first to provide an accurate prediction with good reasons. Many, many people predicted a Trump win, some with good analysis, some with faith, some with hope.

A true establishment clone is the one who says:
Nobody knew...
Everyone thought...
How can they know reality when they are hiding in their safe zone.


25 posted on 11/22/2016 4:11:43 PM PST by spintreebob
[ Post Reply | Private Reply | To 24 | View Replies]


Navigation: use the links below to view more comments.
first previous 1-2021-25 last

Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.

Free Republic
Browse · Search
GOP Club
Topics · Post Article

FreeRepublic, LLC, PO BOX 9771, FRESNO, CA 93794
FreeRepublic.com is powered by software copyright 2000-2008 John Robinson