I am the Danube River.

My spirit is sparkling and swift. I yearn for new experiences and deep connections with people. I’m adaptable, but to a fault; I rarely see danger ahead. I’m capable of infidelity without much remorse. I’m also great at ceramics.

So says Meet Yourself As You Really Are, the oldest, longest, and WTF-est personality quiz I’ve taken. Published in 1936, Meet Yourself is a 336-page home-psychoanalysis test that promises to “‘X-ray’ the reader’s fundamental character.” It does so with an interminable line of questions both probing and random. Are your parents dead? Have you ever had the sensation of standing outside your own body? Do Mickey Mouse cartoons freak you out? What do you think of unskimmed milk?

As you tally “yes” and “no” answers, the book directs you to new sections based on your responses. Somewhere in the middle, you’re categorized as one of 15 rivers—the Nile, Seine, Thames, Missouri, and so on—and eventually you’re offered long-winded personality breakdowns. “As you travel across the network of questions and data by your private track, your story unfolds and your character is explained,” the introduction teases. The book has been described as a Freudian Choose Your Own Adventure, which is accurate enough: It’s like Give Yourself Goosebumps, but instead of escaping the Carnival of Horrors at the end, you learn that you have commitment problems.

Clearly, personality quizzes have some sort of perennial appeal. Facebook newsfeeds are filled with BuzzFeed quizzes and other oddball questionnaires that tell you which city you should actually live in, which ousted Arab Spring ruler you are, and which Hogwarts house you belong in. But these new online quizzes have a dark edge that their analog predecessors didn’t. In the wake of the U.S. election, a secretive data firm hired by Donald Trump’s campaign boasted that it has been using quizzes for years to gather personal information about millions of voters. Its goal: the creation of digital profiles that can predict—and possibly exploit—Americans’ values, anxieties, and political leanings.

Whether this firm, Cambridge Analytica, has actually used predictive profiles to influence people isn’t certain; reports suggest it hasn’t, at least not directly. But the company’s methods nonetheless expose the growing scale of personality analysis online—and the dangers that come with it. On the internet, anything you do is like taking a personality quiz: Everywhere you click reveals something about you. And you’re not the only one who sees the results.

On Meet Yourself’s spine, a silver-painted mirror depicts the personality quiz’s allure: See yourself as you really are. What happens when the mirror is two-way?

* * *

No one seems to know when the personality quiz first gained a foothold in popular culture. The journalist Sarah Laskow has traced its origin in America at least as far back as the late 19th century, “when ladies’ magazines started gaining traction and the yellow press would try anything to sell papers.” But the quiz has persisted with remarkable consistency since, with spikes in popularity during a quick magazine boom immediately post-WWII, the Cosmopolitan quizzes of the 1960s and ’70s, and today’s ubiquitous BuzzFeed quizzes.

This stretch neatly overlaps with the history of the popular quiz’s buttoned-up, high-achieving sibling: the personality test. In-depth psychological assessments like the Myers-Briggs Type Indicator began popping up in the first half of the 20th century for the purpose of scanning and sorting employees in industrial workplaces. While many of these tests, including the Myers-Briggs, have since been dismissed by the scientific community as unreliable—if not dangerously discriminatory—they, too, have persisted, perhaps in part because they at least provide a framework for otherwise-difficult office conversations. Somewhere around 10,000 companies, 2,500 colleges and universities, and 200 government agencies still use the Myers-Briggs in the U.S. today, including the majority of Fortune 500 companies.

Affiliation with these more legitimate-seeming forms of personality analysis has always given the personality quiz a vague air of authority. Indeed, if there’s any one way to characterize quizzes’ mystique, it’s probably that, through all their many iterations, they have somehow managed to tightrope-walk the line between entertainment and science, or at least something approaching science. “BuzzFeed quizzes are crafted to create the illusion of truth, or potential truth,” writes the journalist Caroline O’Donovan, in explaining the fad. She quotes Summer Anne Burton, one of BuzzFeed’s editors: “You sort of write them like horoscopes, with tidbits people can relate to.”

Meet Yourself discovered this strategy long before today’s click-miners. The book’s co-authors, the British novelist William Gerhardie and the Spanish aristocrat Prince Leopold Loewenstein—whose son, incidentally, went on to be the financial manager of the Rolling Stones—flip freely between promises of profound insights and guarantees of fun.

If the limited information out there about the book’s critical reception is any indication, people have never taken Meet Yourself too seriously. A short review in a March 1937 issue of Ohio’s Piqua Daily Call deems it “a fairly amusing way of filling in an odd hour or so,” and includes this sick 1930s burn: “If you sit down prayerfully with it and answer all of its impertinent questions, you will never again be phased by any little thing like an income tax or civil service examination blank.” Decades later, a columnist for The Independent ran into the book on a trip with her boyfriend to her parents’ house, and cracked up when it announced to the family that the boyfriend was “a conqueror of women.”

As BuzzFeed’s quizzes really started gaining steam a few years ago, a deluge of think-pieces attempted to make sense of why people just can’t get enough of them, even when they clearly have little to do with reality. Reasons included narcissism, existential searching, and boredom. Laskow made the case that people simply like talking about themselves. These probably are all true, to some extent. But they overlook something deeper about the nature of personality itself.

* * *

Simine Vazire believes that a good personality test rarely tells you anything you don’t already know. As director of the Personality and Self-Knowledge Lab at the University of California, Davis, she studies how people come to understand who they are. “We know a lot just by being in our bodies, by being ourselves,” she says. Tests promising to unveil hidden truths about their takers—tests known as projective in psychology—are mostly bogus.

What tests offer instead, Vazire suggests, is reflection. “When you have someone summarize to you what you just told them, it gives you a sense of, ‘Oh, yeah, that’s what I was trying to say.’ I think this is just a version of that,” she says. “Because you’re talking about yourself, and you’re answering a bunch of questions about yourself, a test can summarize this information for you. It can give you a precise or better language for summarizing yourself, even if it’s based on what you told it.”

This reflection isn’t the quasi-mystical type of self-knowledge Meet Yourself claims to be after. It doesn’t show you as you really are, but rather helps you articulate who you know yourself to be.

That distinction might sound trivial, but it actually makes a critical point about how personality functions. In its textbook definition, personality is all about patterns: “individual differences in characteristic patterns of thinking, feeling, and behaving,” as the American Psychological Association puts it. Personality, in other words, is not some set thing. It’s the result of a messy web of tendencies and habits, all informed by some incalculable mix of biology, disposition, and learned behavior.

That makes personality slippery. While psychologists debate the granular details of its definition, what it amounts to is that personality is something of a paradox. People tend to have a sense of their own character, but this sense is never complete. We know ourselves, but we don’t.

This is why people love personality quizzes. Beyond vanity and narcissism and harmless fun, taking a personality quiz helps me get out of my own head, to see whether my experience of myself matches up with how others experience me. This is the same reason I sometimes catch myself staring into the mirror even after I’ve double-checked that my fly is zipped and fretted over my oh-so-slightly thinning hair. There’s an element of affirmation, even awe and wonder, to the reflection. That is me in the world.

In this light, personality quizzes actually appear pretty beneficial, or at least innocuous. And maybe they would be, if they left me to my own musings and no one else ever saw the results. But a psychological need for self-reflection gets complicated when the mirror also snatches up information for other people to use.

* * *

In 2008, Michal Kosinski joined a research project that helped revolutionize how data about people is collected. As a graduate student in Cambridge University’s Psychometrics Centre, a department that studies online psychological assessment, he and his classmate David Stillwell distributed a short personality quiz on Facebook that told people how they rated among psychology’s “big five” personality traits: openness, conscientiousness, extroversion, agreeableness, and neuroticism (OCEAN, for short). Upon receiving their results, quiz-takers had the option of sharing their Facebook profiles with Kosinski and Stillwell.

It didn’t take long for the two researchers to amass the largest-ever collection of psychometric scores paired with Facebook profiles. “It was surprisingly easy to get people engaged,” says Kosinski, who’s now a professor of organizational behavior at Stanford’s graduate school of business. “[Stillwell] just sent it to his friends on Facebook. Those friends took the questionnaire, then shared it on their profiles, and suddenly you had thousands of people taking it every day.” Over a few years, Kosinski and Stillwell gathered info from millions of Facebook users.

The two researchers used this data to build a new system for predicting people’s personalities. With access to their subjects’ OCEAN traits and Facebook information side by side, Kosinski was able to correlate what people were like with the personal details available about them online—Facebook likes, gender, age, and so on. Soon, he had an algorithm that, based on analyzing Facebook likes alone, could guess how people think, feel, and act—that messy web of tendencies and habits—with startling accuracy. With 70 likes, the model could predict someone’s personality, as measured by a 100-question personality test, better than that person’s friends could. With 300 likes, it could outperform a husband or wife.

This predictive approach pioneered by Kosinski is what stirred up so much controversy in the U.S. election. Kosinski himself had nothing to do with the data firm hired by Trump’s campaign. Motherboard and Das Magazin have reported that his research appears to have been brought over to Cambridge Analytica by a young colleague of his with ties to the firm’s parent company, the London-based Strategic Communication Laboratories Group, or SCL. (Cambridge Analytica denies that the company or its methodologies have any connection to Kosinski.) After the election, the firm first declared that its brand of “psychographic” profiling played a major role in Trump’s victory, then conceded that it never actually used the approach to influence voters.

Whatever the impact so far, predictive methods like Kosinski’s and Cambridge Analytica’s are pushing open the door to a new—and unsettling—stage of personality analysis. It’s no revelation that personal data can be collected online, and there are already plenty of ways this data can be used to influence people: In politics, for instance, Barack Obama targeted individual voters with psychometrics well before Cambridge Analytica was even formed. But Kosinski saw just how much information can be gained. With personality quizzes, he found what might be the most direct route into people’s hearts and minds yet.

Kosinski never requires Facebook users to give up their profiles to take his quizzes. But the worrying implication of his kind of approach is that online personality analysis can easily blur the line between opting in and opting out. When algorithms can be trained to accurately infer your personality based on anything you do, the internet is a personality quiz—or, at least, it can be, so long as each page visit, web search, and “like” can be gathered and correlated.

Online, before you even click on a quiz, you’re already filling something out.

* * *

There are a lot of good reasons to worry about this technology. Imagine an advertising company that knows you’re self-conscious about your weight, so tries to sell you diet pills. Or—in a hypothetical that often comes up in this kind of discussion—imagine a political campaign that knows you’re prone to anxiety, so targets you with ads about the dangers of the Islamic State. “Big data companies already know your age, income, favorite cereal and when you last voted,” notes a New York Times report on Cambridge Analytica. “But the company that can perfect psychological targeting could offer far more potent tools: the ability to manipulate behavior by understanding how someone thinks and what he or she fears.”

While Kosinski thinks people’s rapidly diminishing privacy online is indeed dangerous, he’s quick to point out that there are potential benefits to personality profiling, too. Targeted ad campaigns could get kids to quit smoking, he suggests. Personalized political messages could inform voters, not pull their strings.

Companies like Cambridge Analytica have a commercial stake in exaggerating their techniques’ reach, as well. “What they’re selling is not exactly snake oil, though it can work as a placebo for panicky candidates who are down in the polls with weeks to go before Election Day,” the journalist Leonid Bershidsky has argued. “But just like artificial intelligence or, say, the blockchain, [data science has yet to produce] killer apps that can ensure a political victory or business success.”

I’m not convinced the nerdy podcasts and obscure track-and-field clubs I like on Facebook will hand the reins of my life to some shadowy corporation anytime soon. But I do think the threat is real—real enough, at least, that I wouldn’t give away my profile information for a personality assessment.

Something Kosinski told me gave me an uneasy feeling I haven’t been able to shake, too. There’s research that has been done on people’s trust in algorithms. A subject talks to an expert on a topic, and the expert offers some sort of insight on that topic, backed by one of two possible justifications: either a) the expert has thought about this for a long time, or b) the expert’s computer calculated the solution. The results show that people are more likely to trust the computer. “We’re being trained by algorithms that they’re always right,” Kosinski says.

Surely, such trust isn’t always misplaced. Vazire, the UC Davis psychologist, admits that she’d probably trust an algorithm over an expert—if she knew the algorithm to be accurate. But what if it’s not? What if, say, it’s built upon data collected by researchers who are prone to error and bias? Or what if it’s intentionally incorrect—sneakily incorrect? Conceivably, an algorithm could know so much about you that it could say exactly what would make you think, act, or feel a certain way.

That’s where the impulse to take a personality quiz keeps me up at night. I’m wired to seek out ways to reflect on who I am, but who I am is slippery—and that makes me open to suggestions. If people’s faith in algorithms continues to grow, it might not be long before I trust a computer to tell me about my personality more than I trust friends or family—or more than I trust myself.

That’s a strange future to imagine. But, hey, I am the Danube River. I’m adaptable. I’m sure I’ll adjust.