Google Knows You Better Than You Know Yourself

Predictive analysis combs through calendars and search histories—and gets in the way of routine self-deception.

Thomas Peter/Reuters

Every weekday at 3:40 p.m., the message popped up on my phone: 22 mins to Work. Google Now was doing its job. It was reminding me where to go and how to get there.

Google Now, in case you don’t know, is the flagship software embedded in the Android smartphone operating system (it’s also available on iOS); its aim is to seamlessly turn your device into a personal assistant that provides “just the right information at just the right time.” It achieves this by using Google’s devouring awareness of your calendar, inbox, and movements. It tracks and predicts your travel, your appointments, your web interests: it tracks you. That may make some people antsy. Not me. Let’s be real: Google knows so much about me already. Might as well put it to use. For me, Google Now is strange and neat and vital.

It can also be kind of a jerk, which brings us back to this daily reminder.

“Work” was the label that, at some point, Google Now had suggested for the place that I went most afternoons: the home of the Kim family, whose kids I tutor. I had okayed this label, tickled that Now was finding patterns. But seeing those daily directions to “Work” quickly began to sting. Because tutoring—and the Kim kids are terrific, really—is not my work. This, what you’re reading, is my work; the novel I’m trying to push up the sales rankings on Amazon is my work. I’m a writer: Writing is my “Work.” Tutoring is a thing I do because, let’s face it, my income as a writer can be inconsistent. Like many writers, I’ve come up with a narrative for this, an equilibration of present and future that looks forward to a time when I don’t need to tutor, sort-of accepts that now is not that time, and holds tight to the fact that my real work is writing—and then there’s this other thing I do for right now that’s really not so bad.

But now, with Now, I had this tool, this digital fifth column, heckling that narrative. Telling me that, inasmuch as “Work” is commonly understood—you go there regularly, to make money—the Kim home was my “Work.” And this sucked because Now was entirely correct. Now’s constant, tabulated observation of me—the truth—was exploding a self-serving conceit I’d constructed. We all have stories we tell ourselves, stories about who we really are. All Google Now does is reveal whether these stories check out.

They often don’t.

* * *

Now’s tracking runs headlong into our need to lie—a little! sometimes!—to ourselves. It’s a truism of this era that Facebook statusing and avatar design and Instagram filters have transformed how we self-present: the way we tell other people true and untrue stories about who we are. What’s transformative about Now is how it makes it harder to tell such stories to ourselves. This matters. Small, self-deceptive fictions are a big part of how we operate. Human beings are not totally awesome at distinguishing between the things we’d like to like and do and the things we actually like and do. So while the sexy endgame of “personal digital assistants” and predictive algorithms may lie in science fictive images of perfected machine intelligence and Scarlett Johansson-y singularities, there are some much more prosaic problems to grapple with right now, today: like the accountability and truthfulness that this technology demands of and imposes on us already.

We’ve been managing this tension for years. Anyone who’s ever cleared a browser history to maintain self-respect, or been appalled by a song that some predictive streaming music service suggests (then ... liked it), has faced technology’s ability to throw us back at ourselves. And even with Now, most revelations feel small. I suppose I do procrastinate by searching movie showtimes, and Chrome on my computer talks to Now on my phone (that’s the point), so, yup, I get notifications about every film screening within three miles of my location pretty much at all times. I asked friends about similar experiences. Most are light: “Now also reminds me *all the time* that I buy my clothing at brooks brothers,” wrote a woman who really doesn’t see herself as a seersucker and polo type. “I'm not entirely comfy with that level of self-awareness. So tracking is disabled,” from an opt-outer. Another friend’s childish Netflix suggestions confirm that, whatever else he might be, he is now a dad. Some start funny, then go melancholy: a buddy whose web browser helpfully suggests a porn site as his likely destination, no matter how many times he Xs out the choice, as if it has “a never-ending supply of my own shame to replenish itself with.”

So, what? What will we gain, or lose, in being stripped of our pretenses? Turns out to be a thorny question. Ask the Internet about “self-deception,” filter the pop psych, and you’re still plunged into an inter- and intra-disciplinary quagmire of psychologists and philosophers disagreeing with themselves and each other on what “self-deception” even means. It exists; it’s important. Everyone’s on board up to that point. But what it is, how it works, and why we do it are topics of active dispute. Is self-deception intentional? What “psychological or temporal divisions” are required to maintain it? Is it simply, in the words of the Stanford Encyclopedia of Philosophy, “motivationally biased belief?”

And, finally, is it good or bad for us?

On this, the evidence is mixed. Evolutionary biologist Robert Trivers argues that self-deception is adaptive, because it makes us better at fooling others. If I really believe I’m a writer, correctly or not, then I’ve a better chance of success when I walk into that meeting with someone who’s looking to hire a writer. This view frames self-deception as a self-fulfilling prophecy, and I’d be very sorry to learn that my smartphone is kneecapping my ability to succeed. I’d be even sorrier, however, to lose my “positive illusions,” as articulated by Shelley Taylor and Jonathon Brown in a landmark 1988 paper. “Most healthy adults are positively biased in their self-perceptions,” Taylor and Brown state. They outline commonly held illusions that fool us into feeling better about who we are, more in control of our lives, and more optimistic about the future. These sound like illusions I need—we all need—to survive.

But self-deception also harms. For one thing, it makes us stupid. In a much-cited 2010 study, researchers discovered that people who performed well on a test because they were given the freaking answer key “interpret their resultant high scores as evidence of superior intelligence,” and that “when asked to predict their performance on a future task, they fail to account for the impact of having had the answers, even when inflated predictions prove costly.” Most of us have spotted this bias in others; most of us are compromised by it ourselves. Even Taylor et al argue for “windows of realism,” observing that people in a decision-making mode tend to be more realistic. And the philosophers point out the ethical importance of self-knowledge, labeling self-deception “a problem of particular concern for moral development, since [it] can make us strangers to ourselves and blind to our own moral failings.”

The issue boils down to this: how is Google Now going to find this balance, for me? How will it temper illusion and reality, according to my complex personal needs? I called Google to ask. Baris Gultekin, one of the founders of and the product lead on Google Now, thinks that the key is in simplifying and refining communications between user and software. “In the real world, you have this communication with your [actual, human] assistant,” he said. “So we want the software to be cognizant of that, to not presume many things, to have a conversation with the user.”

The tension is between seamlessness—a tool that is automatic, predictive, requires no input—and not wanting that tool to be “presumptuous.” This is challenging because all of Now’s data-mining and hey-is-this-what-you’re-looking-for suggestions run on algorithms, and algorithms are algorithmic. Those responses then need to be personalized; to adapt to every individual. “What’s very important is that we need to make sure that the user is in control of all of this,” Gultekin notes. “To create an easy way for the user to communicate with the assistant, to make it for that specific user and no one else.”

I know what this user wants. My perfected digital assistant would help keep me organized, while respecting the touchy stuff. It’d protect healthy fantasies while simplifying my days; allow for “positive illusions” of an aspirational self, present, and future, while making it easier to live effectively in the world as it is. Maybe this is ridiculous—a pipe dream. But hey: 10 years ago, did you even have a smartphone?

The definition of self-deception that I liked best was offered by Chance et al, the researchers with the answer key: “Positive belief about the self that persists despite specific evidence to the contrary.” I can see how this could be dangerous. It also sounds pretty nice. After a few months, I got fed up with my phone telling me what I did, and therefore who I was. I wiped the Kims’ address from Now, told it I was “no longer interested” in travel to that place. I kept going, of course, and so Now kept asking about it; without a name for the location, it would just pin the address: street name and number. Impersonal. Almost a slight, against a family I like and a part of my life that is ... part of my life. I finally came up with a solution:

22 mins to the Kims

Maybe someday, Google Now would’ve known to do this from the start. I still kind of wish it reminded me less often, though it’s still mostly right when it does. And the label is both true, and acceptable. I can live with that. For now.