Stressed, Persuasive and Anxious: We Used a Hollywood Script Analysis Tool to Analyze the Debate

Stressed, Persuasive and Anxious: We Used a Hollywood Script Analysis Tool to Analyze the Debate

CNN reporter Dana Bash called Tuesday's presidential debate a "shit show," NBC's Andrea Mitchell called it a "disgrace" and ABC's Martha Raddatz called it "mud wrestling in the middle of a pandemic."

But Yves Bergquist, founder of CortoAI, saw it as an opportunity to show how AI-powered software used for Hollywood scripts could reveal a candidate's underlying emotions, not always evident on camera.

"The words we use are indicative of our emotions and perceptions of social dynamics," Bergquist told dot.LA.

Corto's software analyzes texts and gauges the extent to which any one of 59 "emotional tonalites" are present, on a scale of 0-100. Anything above 75 is considered particularly salient, while values below 25 convey a notable absence.

Some of what Corto's debate analysis reveals is unsurprising. It found both candidates used language that was stressed, persuasive and anxious. Neither conveyed much happiness.

But Bergquist was impressed by how Trump's high-scoring traits mirrored a typical Hollywood drama lead. The AI "seems to confirm that Trump is being a president in character," Bergquist said.

Biden, meanwhile, "has a much more real-person personality."

Somewhat surprisingly, Corto found the former VP considerably more assertive and self-assured than Trump, while Trump was more trusting and imaginative.

"I don't want to create the impression that this is perfect science," Bergquist said. "This is really powerful but it's not a perfect tool. It's really important to remind people that what this is, is a good tool that in some ways is more nuanced than the human mind and in some ways is less nuanced."

Since 2016, Corto has worked symbiotically with the USC's Entertainment Technology Center, a think tank founded in 1993 by Star Wars creator George Lucas to support cooperative technological advancement in Hollywood. It is currently used by a variety of studios to greenlight scripts, Bergquist said. Producers use the tool to analyze whether a character's tone is hitting the mark.

The software learned to link words to emotions through a combination of text-based training and guidance from professional psychologists and psychiatrists.

Corto does not pick up on body language or tonal cues — just the words. This, however, can reveal some intriguing, between-the-lines insights.

For example, "when Corto sees aggressive behavior it also sees a lot of melancholy. And that makes sense: a lot of times when people are being aggressive they're in pain," Bergquist said.

Many who watched the debate can probably relate to that pairing.

Below are each of the candidates' ten highest and five lowest emotional tonalities according to Corto, and the five areas with the greatest difference between the two.

Trump's Top Ten

Persuasive: 95.38

Stressed: 94.23

Type A: 91.1

Depression: 88.3

Artistic: 83.85

Melancholy: 83.53

Cautious: 79.13

Neurotic: 78.82

Sociable: 78.19

Anxious: 76.7

Biden's Top Ten

Money Oriented: 90.56

Stressed: 89.03

Persuasive: 84.23

Anxious: 78.77

Insecure: 78.34

Disciplined: 78.06

Assertive: 78.06

Conscientiousness: 77.9

Melancholy: 77.01

Cautious: 75.36

Trump's Bottom Five

Happiness: 8.71

Self Assured: 16.28

Liberal: 18.64

Impulsive: 19.25

Cold: 20.39

Biden's Bottom Five

Impulsive: 18.17

Happiness: 20.41

Genuine: 24.41

Imaginative: 24.97

Emotionally Aware: 25.07

Largest Differences

Family Oriented: 38.84 (Biden: 62.48; Trump: 23.64)

Self Assured: 26.1 (Biden: 42.38; Trump: 16.28)

Trusting: 22.47 (Trump: 70.74; Biden 48.27)

Imaginative: 20.36 (Trump: 45.33; Biden 24.97)

Religion Oriented: 19.28 (Biden: 56.49; Trump: 37.21)

---

Sam Blake primarily covers entertainment for dot.LA. Find him on Twitter @hisamblake and email him at samblake@dot.LA

Subscribe to our newsletter to catch every headline.

Minutes into filling out my absentee ballot last week, I was momentarily distracted by my dog Seamus. A moment later, I realized in horror that I was filling in the wrong bubble — accidentally voting "no" on a ballot measure that I meant to vote "yes" on.

It was only a few ink marks, but it was noticeable enough. Trying to fix my mistake, I darkly and fully filled in the correct circle and then, as if testifying to an error on a check, put my initials next to the one I wanted.

Then I worried. As a reporter who has previously covered election security for years, I went on a mini-quest trying to understand how a small mistake can have larger repercussions.

As Los Angeles County's 5.6 million registered voters all receive ballots at home for the first time, I knew my experience could not be unique. But I wondered, would my vote count? Or would my entire ballot now be discarded?

Read more Show less

When the founders who lead the ten young startups selected for the 2020 Techstars LA class begin their three month accelerator program Monday, they won't be gathering in the Mid-Wilshire office and shaking hands as every other class has done. Like the rest of us, they will be working at home because of the coronavirus. Dinners, meetings, socializing, and mentoring sessions will all be online.

"A big part of the magic of the program is the relationships that are from proximity and from everyone working together in the same space and so what we're doing is we're endeavoring to create as much as that connection in the virtual world as possible," said Anna Barber, managing director of Techstars LA.

Read more Show less
RELATEDEDITOR'S PICKS

Trending