Friday, June 24, 2016

"Whom are you voting for? This guy can read your mind." -- with that headline, what could go wrong?



A few quick thoughts on this WaPo piece by Jacob Bogage:


Neuroscientist Ryan McGarry swabbed a brain activity headset with saline solution and lowered it onto Brian Hazel’s head, connecting circular prongs gingerly to different spots on his skull.

Then he showed Hazel 40 minutes of presidential debates and commercials as Spencer Gerrol tried to read his mind. Which of the candidates did Hazel, an African American real estate professional in his 40s, support?

Gerrol, founder and chief executive of creative agency Spark Experience, based in Bethesda, Md., stared through a one-way mirror from an observation room. “Trump,” he said five minutes into Hazel’s session.

Gerrol was right. “Donald Trump,” Hazel said after his session. “I’m all about jobs, and if I look at the whole field and who has created the most jobs, I think he would be able to do that better than anyone else.”

Spark's experiment is leading a new method of human factors research, which asks test subjects to interact normally with everyday products or services while researchers track their emotional reaction and attentiveness. The agency gathers data from four tests — electroencephalograms, galvanic skin responses, eye tracking and microfacial recognition — to instantaneously determine which candidate a subject supports, down to the severity of emotional response.

The firm gathered more than 30 test subjects from across the political spectrum in May. Their findings, released last week, show that what people feel and what they say they feel are rarely the same.

The experiment might be the closest the country gets to an explanation of this crazy presidential campaign without dissecting a brain.

Instead of placing participants in focus groups or asking them to fill out a questionnaire, experimenters harvest data directly from the brain using technology called "BrainWave." Companies such as Nielsen and Affectiva also conduct similar testing, which can gauge the effectiveness of advertisements and viewer response to TV shows and movies.

“This isn’t science fiction,” Gerrol said. “I can’t read your thoughts, but I can read your emotions.”

...

But it has particular application this election cycle because voters' responses have been so unpredictable. Spark's study may have provided some answers, namely that test subjects may subconsciously lie when they tell people what political messaging works and what doesn't.

“If you ask somebody what they’re feeling, you're not going to get a very accurate response. Emotion is subconscious,” Gerrol said. “The idea is to measure emotion without asking.

“When people make decisions, it’s tempting to think there’s a lot of rational thought. That’s not really true. You can't have decisions without emotion.”

...

Scores of political prognosticators predicted Trump’s demise almost a year ago. They predicted that Democratic candidate Bernie Sanders would fizzle out.

Alas, no political prediction has rung true, leaving political professionals and junkies alike asking, “What gives?” and “What are Trump and Clinton voters thinking?”

1. First and most obvious, a minimum level of skepticism is required in these situations. None is to be found in the article.

2. There are huge causal and logical leaps here. Determining who a debate viewer supports based on biometrics shouldn't be that difficult -- hell, you could probably do a good job just watching facial expressions -- but Gerrol claims all sorts of insights about the subjects' decision-making process.

3. All based on 30 subjects. If you read the article, you'll see that Gerrol seems able to infer some important relationship from almost every line of the debate. That's a huge edifice for such a narrow base.

4. Let's stick with the 30 a bit more. There's every reason to believe we are traversing a rugged landscape here, with a definite possibility of interactions and non-linear relationships. If you have to disaggregate 30 subjects by gender, age, etc., you start getting some very small numbers.

5. Why do they always want to do politics? There are lots of choices (sometimes strongly held -- "Coke or Pepsi?") that are difficult to explain and predict, choices that are clearly emotional and often marketing driven. Voting is not one of those. We have pretty good models to explain the process which usually comes down to rational factors like social issues and economic interests. If you factor in what information the voters rely on, the choice becomes even more rational (if you believe what you see on Fox News, supporting Trump is a sensible decision).

6. As for this cycle being unpredictable... A lot of people in the journalistic establishment don't want to admit this but the people who got this campaign humiliatingly wrong did so because they argued against the data (see here and here). As far as I know, actual political scientists did fine.

7. Spencer Gerrol is a TEDx presenter.

No comments:

Post a Comment