I Asked A Computer To Be My Life Coach

Dec 22, 2015
Originally published on December 28, 2015 5:57 pm

Editor's Note: This post accompanies a story that you can hear on the NPR One app by following this link.

The words you use betray who you are.

Linguists and psychologists have long been studying this phenomenon. A few decades ago they had a hunch that the number of active verbs in your sentences or which adjectives you use (lovely, sweet, angry) reflect personality traits.

They have painstakingly pinpointed various insights. For example, suicidal poets, in their published works, use more first-person singular words (like "me" or "my") and death-related words than poets who aren't suicidal. People in positions of power are more likely to make statements that involve others ("we," "us"), while lower-status people often use language that's more self-focused and ask more questions. Comparing genders, women tend to use more words related to psychological and social processes, while men referred more to impersonal topics and objects' properties.

(This 2010 paper in the Journal of Language and Social Psychology goes into great detail about the "psychometrics" of words.)

This research suggests that Internet companies such as Facebook and Google, with their troves of written expressions, are sitting on powerful insights about us as people. But if you ask them, "Hey, can you give me the take on me that you've got in-house or that you've built for advertisers, with my anonymized data?" — they won't give it to you. I actually did ask, and they don't have that kind of offering.

But I've found someone who does: IBM's Watson division. Researchers there have taken the personality dictionaries already created by scientists, dropped them into Watson (the computer that won Jeopardy!), and sent it off to apply it to people on Twitter, Facebook, blogs. That forms a digital population of people and personality types. Over time, more text from more people will help Watson get smarter. (Yes, this is machine learning.)

In its own studies, IBM found that characteristics derived from people's writings can reliably predict some of their real-world behaviors. For instance, people who are less neurotic and more open to experiences are more likely to click on an ad, while people who score high on self-enhancement (meaning, seek personal success) like to read articles about work.

For IBM, these kinds of interpretations can become a business opportunity.

Understanding people to sell them things is obviously a very big business for marketers. IBM senior researcher Rama Akkiraju suggests other uses: by public relations firms looking for journalists who sound friendly on a specific topic; by editors who want their writers to set a certain tone; by employers looking for a worker who fits their corporate culture.

"We're moving to make it easier for people to consume insights," she says.

This use of Big Data, of course, raises serious privacy concerns, which we have examined in many stories. In this exploration, I decided to take a deep dive into Watson's personal insights — what they can teach me about my career choices and my love life (yep, really went there).

You can listen to my story on NPR One.

Not all of the tools I used are publicly available, but you can try out a couple of them. Click on "view demo" here to test the tone analyzer that evaluated the tone expressed in my love letters. And here's a tool to analyze personality through writing.

For now, Watson's personality analytics is a work in progress and not easy on the eyes. The pie chart it spits out from a person's social media posts, which you saw above, is a messy hodgepodge of about 50 traits. Plus, given how people curate their digital presence, the words we use online may be a highly biased indicator of who we are.

"It's very interesting as a general curiosity," says Sina Khanifar, a San Francisco-based technologist, "but what would really get me excited is if it made a particular recommendation."

Khanifar says many tools exist to help you quantify yourself, track your running speed or breathing patterns. What few of them do is actually suggest how to improve your life. And, he says, people don't just want to pay for insight. "When you go to see a therapist, it is about self-knowledge. But it's also about a change."

A friend recently mused about what this kind of tool could do for dating. People lie about themselves on dating sites chronically. What if Google developed a service to mine your mail and search and paired you with the perfect partner?

That could be amazing, or amazingly creepy.

Copyright 2018 NPR. To see more, visit http://www.npr.org/.


Ever wonder what your life online reveals about you IRL - in real life? Your Facebook posts, your Gmail, what do they reveal about who you are? That's the subject of this week's All Tech Considered.


CORNISH: Your digital life leaves a trail social media companies are already following, so imagine if you could mine your own accounts for insights into your personality. NPR's Aarti tried to do just that.

AARTI SHAHANI, BYLINE: I'm going to walk you through how I mined my own data and what it taught me about my career choices and my love life - yep, really going there. If you ask Facebook or Google hey, give me the take on me that you've got in-house or that you in share different ways with advertisers, they're not going to give it to you. I did ask. They said no, they don't have that kind of offering. But another company recently came on my radar that could kind of get at the same thing, IBM.


UNIDENTIFIED MAN: With Watson analytics, you can instantly access and use...

AARTI SHAHANI: ...Specifically, it's the Watson division. IBM took the most famous computer on Earth, the computer that won "Jeopardy," and retooled it. Instead of answering trivia about the world, Watson is now parsing through people to give personal insights.


UNIDENTIFIED MAN: What are you waiting for? Just add your data and go.

AARTI SHAHANI: Here's how it works. A few decades ago, linguists and psychologists had a hunch that the words you use - the number of active verbs in your sentences, the adjectives like lovely, sweet, angry- they reveal your personality. It turns out, says IBM analyst Rama Akkiraju...

RAMA AKKIRAJU: ...That unconsciously, the choice of words that we make as we express ourselves truly does reflect who we are. And those things come out in some way or the other, no matter which medium.

AARTI SHAHANI: The scientists went through the painstaking task of creating personality dictionaries, correlating words in English to five major characteristics. You can remember it as OCEAN - O for openness, C for conscientiousness, E for extroversion, A for agreeableness and N for neuroticism. OCEAN. IBM took those dictionaries, dropped them into Watson and said go apply that to people on Twitter, Facebook, bloggers to create a digital ground floor of personalities.

AKKIRAJU: We took your Facebook posts...

AARTI SHAHANI: ...Facebook has this little hidden option to download your entire profile, so I did and sent it over. Watson spit out my personality pie chart, who I am according to what I post. It's 50 traits, each with a percentile score, kind of like the SAT, ranking you compared to others. I shared it with the human being who knows me best.

ANGELLY SHAHANI: It was overwhelming. It was - like, it's very colorful.

AARTI SHAHANI: My big sister, Angelly Shahani, or Ang.

ANGELLY SHAHANI: And there's a lot that jumps at you.

AARTI SHAHANI: Ang and I parse through the chart. On openness and adventurousness, I rank high. On modesty, I rank rock-bottom. She agrees with all that.

ANGELLY SHAHANI: If I had to give you numbers on a lot of these, it'd probably close.

AARTI SHAHANI: A few things she found to be wrong.

ANGELLY SHAHANI: Truth on this is kind of low, 18 percent.

AARTI SHAHANI: Now it is possible to be very different in different places, at home versus a cocktail party versus Facebook. And when I think about, it is true that on Facebook I talk about sad things, like in the news, much more than I do in person. And I almost never crack jokes there.

ANGELLY SHAHANI: Right. So you're one of those people who don't put their best (laughter) their happy Facebook self forward. Your personality on Facebook is sadder (laughter).

AARTI SHAHANI: On self-transcendence, meaning how much I show concern for the welfare of others, I rank in the middle.

ANGELLY SHAHANI: I mean, it might've been way higher when you were in your early twenties, but I think time has probably kind of beaten that - kind of moderated you on that.

AARTI SHAHANI: She hit a nerve. Before I became a tech reporter in Silicon Valley, I had an entire other life that was all about caring for others. I spent nearly a decade in prisons trying to help immigrant families who were being deported. I look at this score and another from Watson, sympathy. It's my single highest score, 92nd percentile. And I ask my big sister, is there a glaring disconnect between what I've chosen to do and who I really am?

ANGELLY SHAHANI: I just don't think you're thinking about it correctly.

AARTI SHAHANI: Angelly points out the top three traits.

ANGELLY SHAHANI: Those top three things - sympathetic, challenges authority and is very intellectual. So - I mean, when you think about it, doesn't that make journalism a great career for you?

AARTI SHAHANI: Watson doesn't answer that question, but it does give you a new tool to dig in. There's another tool as well, an emotions analyzer - really. Something to tell you not your personality, but the emotions you're conveying in a given moment. I decided to put some love letters into it, writing to and from my ex-boyfriend.

I wrote to my ex-boyfriend - when we were still together, this is something I wrote to him, and he wrote a response.

And I brought a human expert into the mix, Fiachra Figs O'Sullivan, a leading couples' counselor in San Francisco. I want Figs, as he's known, to read the e-mails and appraise Watson's take.

O'SULLIVAN: Yeah, let me see.

AARTI SHAHANI: I fell for my ex through letters. Early on, we wrote to each other a ton. But soon after we got into a committed relationship, his expressiveness, it evaporated. I wanted to fix that. Figs reads the letters and then I ask him a question that's a bit of a setup.

Are either or both of us angry?

O'SULLIVAN: No. You both actually do an amazing job, which is very commendable, at talking primarily about yourself.

AARTI SHAHANI: Meaning we start sentences with I feel or I think, not with insults like you suck or you do this to me. So his read makes sense. IBM Watson, however, has a very different read.

O'SULLIVAN: It says that you're 100 percent angry.



AARTI SHAHANI: ...It is true, Figs says, that we each gave hints of anger, disappointment, impatience. Those weren't the majority of lines, but it was there. Watson even locates the line, the specific line in my letter that has the most anger.

O'SULLIVAN: So let me have a look at this line. Where is - so this is line two. (Reading) It's been a while since we've properly written to each other. This is interesting because you even have the smiley face, right?

AARTI SHAHANI: I thought the line sounded like it's been a while, my love, since we've written properly. Watson seems to think it's more like I've got my finger tapping the table. It's been a while, come on now. Watson also says I was off the charts cheerful, masking my anger with sweet words. I wonder if it would've been helpful to use a tool like Watson before hitting that send button. Figs thinks it would've just helped me be manipulative or even prosecutorial.

O'SULLIVAN: You take your ex-boyfriend's e-mail and you run it through and you show him - see, see? Even Watson says that you have a hard time sticking with things.

AARTI SHAHANI: If you could download an app to see your digital personality, would you? Using Watson, I found myself sucked in. I got knots in my stomach, not unlike when you're sitting at the doctor's office. The difference is doctors can presumably help if something's wrong. This analysis doesn't come with a clear prescription. Maybe that's why the internet companies collecting our data don't show it to us. Aarti Shahani, NPR News, Silicon Valley.


CORNISH: That's All Tech for this week, but there's always more on our blog. Go to npr.org/alltech and we've got a little treat for you there. You can test your knowledge of some of the quirkier tech stories we aired this year. It's a quiz. Don't worry, it's multiple-choice. Here's a sample question. What animal did scientists at Texas A&M University try to transform into a cyborg - a spider, a fly, an ant or a cockroach? Or this question - what punctuation in a text message have researchers found make you come off as insincere? Take the quiz, find the answers at npr.org/alltech. Transcript provided by NPR, Copyright NPR.