Yes, We Do Know What You're Thinking

I sometimes hesitate to tell people I'm a psychologist because it can put them on guard. I've heard all of the following reactions:

"Oh no! You'll be able to tell how crazy I am!"

"Can you tell what I'm thinking?"

"Are you analyzing me now?"

I quickly clarify that I'm not that type of psychologist. I'm a cognitive psychologist who studies basic reasoning, not a clinical psychologist or a therapist.

Yet the idea that a total stranger could instantly "read" you, however unlikely, is sufficiently credible that we're willing to entertain characters like Gregory House on House, Cal Lightman on Lie to Me and, of course, Sherlock Holmes in contemporary manifestations, such as Sherlock and Elementary. These masterminds can tell, after casual inspection, how your love life is going, the state of your finances and what you had for breakfast.

Are these inferences a violation of privacy?

On the face of it, the answer is "no." Such inferences, however uncanny, are based on information you've made public in the way you present yourself. They're nothing like the violations of privacy that we worry about with NSA snooping, social network settings or the Heartbleed Bug that's got us all worried.

And yet ... they go well beyond what you intended to disclose.

Questions about how we learn about others, and what we inadvertently reveal about ourselves, have been on my mind after reading about a neat new paper by Dan Jurafsky and colleagues, published last week in the journal First Monday.

Using data-mining techniques, the researchers studied 900,000 online restaurant reviews and discovered some interesting relationships between the reviews and their writers. As the headline on a Stanford News report boldly summarized: "Online Food Reviews Reveal The Inner Self."

Quoting the Stanford News report, here were some of the main findings:

"Positive reviews of expensive restaurants tended to use metaphors of sex and sensual pleasure, such as 'orgasmic pastry' or 'seductively seared foie gras.'"

"Positive reviews of cheap restaurants and foods often employed metaphors of drugs or addiction — 'these cupcakes are like crack.'"

"Women were more likely than men to use drug metaphors to describe their attitudes toward food."

"The foods most likely to be described using drug metaphors were pizza, burgers, sweets and sushi."

In tweeting that you need a hit of chocolate, are you inadvertently communicating something about your socioeconomic status? Or your gender?

I mention this study as a lighthearted illustration of a much more general point: that with the rise of big data and powerful methods for data analysis, those with access and know-how will increasingly be able to identify subtle associations between public bits of information and their individual human producers. And that, in turn, can support reliable inferences about those individuals from public information.

When — if at all — do such efforts cross the line from merely socially inappropriate and embarrassing, in a Dr. House or "Holmesy" sort of way, to worrisome violations of privacy? Or do they just feel that way? And if so, why?

As a psychologist, I don't know what you're thinking. But it's quite possible that some of the folks with access to loads of data and processing power have an inkling.

You can keep up with more of what Tania Lombrozo is thinking on Twitter: @TaniaLombrozo

Copyright 2018 NPR. To see more, visit