Crisis Text Line has received five million texts, providing a unique corpus of data.Illustration by Cristiana Coucerio; photographs: Top to Bottom: Nick Vaccaro / The Image Bank / Getty; Anna Eckold / Moment / Getty; Atsushi Yamada / Photodisc / Getty

In 2011, a young woman named Stephanie Shih was working in New York City at DoSomething.org, a nonprofit that helps young people start volunteer campaigns. Shih was responsible for sending out text messages to teen-agers across the country, alerting them to various altruistic opportunities and encouraging them to become involved in their local communities: running food drives, organizing support groups, getting their cafeterias to recycle more. Silly, prankish responses were not uncommon, but neither were messages of enthusiasm and thanks. Then, in August, after six months on the job, Shih received a message that left her close to tears for the rest of the day. “He won’t stop raping me,” it said. “He told me not to tell anyone.” A few hours later, another message came: “R u there?” Shih wrote back, asking who was doing this. The next day, a response came in: “It’s my dad.”

DoSomething.org had no protocol for anything like this, so Shih texted back with the contact information for RAINN (Rape, Abuse & Incest National Network*)*{: .small}, the country’s largest anti-sexual-assault organization. But the texter indicated that she was too scared to make a phone call. “This is the right thing to do,” Shih insisted. There was no reply. “Not knowing if she was safe or had gotten help or would ever get help consumed my thoughts,” Shih told me last fall. She printed out the text messages and handed them to her boss, Nancy Lublin, DoSomething.org’s C.E.O.

“I’ll never forget the day,” Lublin said. “It was like I’d been punched in the stomach.”

That week, Lublin and Shih started work on what two years later became Crisis Text Line, the first and only national, 24/7 crisis-intervention hotline to conduct its conversations (the majority of which are with teen-agers) exclusively by text message.

Depression is common among teens, and its consequences are volatile: suicide is the third leading cause of death for Americans between the ages of ten and twenty-four. In that same age group, the use of text messaging is near-universal. The average adolescent sends almost two thousand text messages a month. They contact their friends more by text than by phone or e-mail or instant-message or even face-to-face conversations. A. T. & T. offers parents a tutorial in deciphering acronyms used by children (PIR stands for “parent in room”). For teens, texting isn’t a novel form of communication; it’s the default.

People who spent their high-school years chatting with friends on landlines are often dismissive of texting, as if it might be a phase one outgrows, but the form is unparalleled in its ability to relay information concisely. The act of writing, even if the product consists of only a hundred and forty characters composed with one’s thumbs, forces a kind of real-time distillation of emotional chaos. A substantial body of research confirms the efficacy of writing as a therapeutic intervention, and although tapping out a text message isn’t the same as keeping a diary, it can act as a behavioral buffer, providing distance between a person and intense, immediate, and often impulsive feelings. Communication by text message is halting and asynchronous, which can be frustrating when you’re waiting for a reply but liberating when you don’t want to respond. The young people who contact Crisis Text Line might be doing so between classes, while waiting in line for the bus, or before soccer practice. In addition, more than ninety-eight per cent of text messages are opened; they are four times more likely to be read by the recipient than e-mails. If you’re a parent, you know that, even if your son does not text you back about where he is, he has read your message. If you are a distressed teen or a counsellor, you know that what you say will be read.

A person can contact Crisis Text Line without even looking at her phone. The number—741741—traces a simple, muscle-memory-friendly path down the left column of the keypad. Anyone who texts in receives an automatic response welcoming her to the service. Another provides a link to the organization’s privacy policy and explains that she can text “STOP” to end a conversation at any time. Meanwhile, the incoming message appears on the screen of Crisis Text Line’s proprietary computer system. The interface looks remarkably like a Facebook feed—pale background, blue banner at the top, pop-up messages in the lower right corner—a design that is intended to feel familiar and frictionless. The system, which receives an average of fifteen thousand texts a day, highlights messages containing words that might indicate imminent danger, such as “suicide,” “kill,” and “hopeless.”

Within five minutes, one of the counsellors on duty will write back. (Up to fifty people, most of them in their late twenties, are available at any given time, depending upon demand, and they can work wherever there’s an Internet connection.) An introductory message from a counsellor includes a casual greeting and a question about why the texter is writing in. If the texter’s first message is substantive (“My so-called boyfriend is drunk and won’t stop yelling at me”), the counsellor echoes the language in order to elicit additional details (“I’m so sorry to hear that. Can you tell me a little more about what your so-called boyfriend is saying?”). If the incoming message is vague (“Life sucks. I’m freaking out”), the reply will be more open-ended, while gently pressing for greater specificity (“So what’s going on tonight?”). An average exchange takes place over a little more than an hour, longer if there is the risk of suicide.

Counsellors are trained to put texters at ease and not to jump too quickly into a problem-solving mode. Open-ended questions are good; “why” questions are bad. Also bad: making assumptions about the texter’s gender or sexual orientation, sounding like a robot, using language that a young person might not know. Techniques that are encouraged include validation (“What a tough situation”); “tentafiers” (“Do you mind if I ask you . . . ”); strength identification (“You’re a great brother for being so worried about him”); and empathetic responses (“It sounds like you’re feeling anxious because of all these rumors”). The implicit theory is that in a conversation people are naturally inclined to fill silences.

It is important to type carefully. In text messages sent to friends, typos can be an indication of intimacy. But a typo appearing on the cell-phone screen of a distressed teen-ager can undermine the sense of authority he’s looking for. “You have to train yourself not to hit that return button automatically,” a sixty-year-old counsellor from California told me. (Crisis Text Line counsellors are free to give a real or assumed first name to people who text in.) It is also regarded as a mistake to embrace teen-age patois too enthusiastically. One volunteer told me that she tries not to use acronyms. “I sometimes worry that it would come across as too ‘Oh, I got you!’ ” she said. Neutral language allows the texter to feel anonymous. These people have contacted a stranger for a reason. They aren’t looking for friendship.

Often, the conversations are about minor-seeming problems—fights with friends, academic pressure from parents—and the bar for helpfulness is quite low. “A lot of times, when chatting with young people, it’s clear that they just need someone to listen to them,” one counsellor told me. “Sometimes it’s obvious. They’ll say, ‘Thanks for listening. Nobody ever does that,’ and at other times it’s less explicit; they just want to get everything out, and they provide you with a very, very detailed account.”

The etiquette encouraged for counsellors can be surprising. When an agitated friend texts me bad news (a breakup, a layoff, a sudden rent increase), my instinct is to find a positive response to the predicament (“But you didn’t even like him!” “Now you can finally go freelance!” “MOVE!”). But this is precisely what one is not supposed to do when communicating with a teen-ager in crisis. Instead, counsellors are trained to deploy language that at first seems inflammatory: “You must be devastated” is a common refrain; so is “That sounds like torture.” The idea is to validate texters’ feelings and respond in a way that doesn’t belittle them.

Thomas Joiner, a psychology professor at Florida State University and one of the country’s leading suicide experts, pointed out another way in which conversational norms can be counterproductive. “From a clinical standpoint, one common misstep is tiptoeing around issues and treating them like taboos,” he said. “It sends the implicit message that it’s really not O.K. to talk about it, and if the counsellor doesn’t feel comfortable talking about it why would the teen-ager?” It takes practice to tell someone who is suffering that he has a real problem, and that, though things may get better, it may not be anytime soon.

Each day, on average, Crisis Text Line instigates at least one active rescue of a texter who’s thought to be in immediate danger of suicide. During active rescues, the counsellor asks questions as casually as possible—Are you alone? Do you have someone you trust whom you want us to contact? Is your door locked?—and feeds the answers to a trained supervisor, who, in turn, contacts the police. One counsellor, a twenty-eight-year-old former Division 1 basketball player who began volunteering last May, told me that the very first conversation he took was with a teen-ager who was contemplating jumping off a roof. The exchange lasted for an hour, and, by the end of it, the teen was in the car with a parent, driving to the hospital.

In 1906, a woman staying at a Manhattan hotel asked the manager if she could speak to a minister. The manager tried calling Harry Marsh Warren, a minister at a Baptist church, but was unable to get through. The following morning, the woman was found unconscious beside a bottle of poison and was rushed to Bellevue. Warren visited her as she lay dying, and she told him, “I think maybe if I had talked to someone like you I wouldn’t have done it.” Soon after, Warren started the Save-A-Life League, the country’s first suicide-prevention organization. He placed an ad in a local paper, encouraging anyone contemplating suicide to contact him. News spread quickly, as did the organization’s reach. By the time Warren died, in 1940, there were branches across the United States, as well as in London and Paris, and the league was helping around a hundred people each week—providing counselling, free hospital beds, and legal services. It also raised summer-camp tuition for the children of suicides.

In the nineteen-forties, a Boston-based psychiatrist named Erich Lindemann attempted to make the field of crisis intervention more empirical. He conducted his research in the wake of the Cocoanut Grove disaster, of 1942, a fire at a Boston night club that killed nearly five hundred people. Lindemann interviewed dozens of survivors and published a paper based on his findings. He determined that people in crisis are open to help, and that appropriate and expedient treatment could avert the need for long-term psychotherapy, which was the leading method of mental-health treatment at the time.

A decade later, in London, Chad Varah, an Anglican priest, founded a suicide hotline in the crypt of his church: the Samaritans took its first call on November 2, 1953. The idea for the service had come to Varah when he held a funeral for a fourteen-year-old girl who had killed herself on getting her first period, which she thought was symptomatic of a sexually transmitted disease. In 1958, the Los Angeles Suicide Prevention Center was founded, the first in America. It formulated a set of protocols that were adopted by other centers in the United States and abroad. Volunteers establish rapport, define the problems, and assess the risk of self-harm. They aim to reduce anxiety, discussing how callers have coped with similar problems in the past. Finally, they develop a specific plan of action.

In 1963, President Kennedy signed the Community Mental Health Act into law. Among other things, this legislation provided federal funding for community-based mental-health-care centers. Crisis-intervention hotlines soon proliferated, with separate lines for those experiencing suicidal thoughts, drug addiction, sexual abuse, eating disorders, and so on. This arrangement insured that callers would talk to someone who had a reasonable degree of expertise in what was troubling them. Crisis Text Line departs from this practice; there’s just one number to text, whatever ails you. The medium makes it easy for volunteers to look up information, and the C.T.L. interface enables them to enlist the help of colleagues who have training in a particular area. Nancy Lublin often explains the system by saying, “People don’t experience life in an issue-specific way.”

“Forget it, Jake. It’s Funkytown.”

Texting has other advantages. The fact that counsellors can work from home while eating Chinese takeout—and can even trade shifts with one another—makes it easier to attract volunteers. More important, from an adult perspective teen-agers can often seem willfully uncommunicative in speech but are forthcoming, even garrulous, when texting. “On the phone, you have to ask a few more questions, sort of explore a little bit more to find out what’s really going on,” Jen James, who works for C.T.L. in Michigan, told me. “With the text line, they are pretty open. They just come out and tell you and want to talk about it.” Research bears out this observation. According to Fred Conrad, a cognitive psychologist and the director of the Program in Survey Methodology at the University of Michigan Institute for Social Research, people are “more likely to disclose sensitive information via text messages than in voice interviews.” To those who didn’t grow up texting, this seems counterintuitive. Texts are a written record, after all, and what if the wrong person saw them? But, in practical terms, text messaging affords a level of privacy that the human voice makes impossible. If you’re hiding from an abusive relative or you just don’t want your classmates to know how overwhelmed you feel about applying to college, a text message, even one sent in public, is safer than a phone call. What’s more, tears go undetected by the person you’ve reached out to, and you don’t have to hear yourself say aloud your most shameful secrets.

It’s difficult not to notice the merriness of the place in which these dismal matters are analyzed. DoSomething.org’s headquarters, where the employees of Crisis Text Line also work, is situated just north of Union Square, in New York City. Stray balloons cling to the ceiling; there’s an aquarium and a disco ball. Many of the staff members—eighty people, of whom only fourteen are over thirty—seem to spend much of the day without shoes. Under the bright lights and amid the cheerful buzz of people born after the Gulf War, one has a sense of observing kids collaborating on a group project at a school in a county with high property taxes.

DoSomething.org and Crisis Text Line are separate entities, but Nancy Lublin is the C.E.O. of both. She is forty-three, and likes to refer to herself as the C.O.P. (Chief Old Person). She speaks quickly, in a frank but friendly tone, and is unafraid to contort her face into goofy, sometimes even self-consciously grotesque expressions. Lublin brags on behalf of her employees, often in their presence, and has the air of a beloved social-studies teacher who swears. When we first met, last October, she was wearing a cotton scarf printed with the face of Hello Kitty.

Lublin grew up in Hartford, Connecticut, a city that she disdainfully identifies as “the insurance capital of the world.” She attended Brown University and then Oxford, where she was a Marshall Scholar studying political theory—or, as she puts it, “a lot of dead white guys.” Upon returning to the States, Lublin enrolled in law school at N.Y.U. “My whole life people were, like, ‘Oh, you have a lot to say! You’re going to be a lawyer,’ ” she told me. But Lublin hated “everything from nine-point font in thick textbooks to the Socratic method to classmates who were really just fighting for the right to be on law review, which was looking up your professor’s footnotes for an article that was going to appear in a journal that maybe twelve people would read.” She dropped out after her fourth term.

While still at N.Y.U., Lublin recalled something that her father, a lawyer, once told her about his method of hiring new secretaries for the firm. Looking out his window, he would watch the women walk from their car to the front of the building and know before they reached the door whether he would hire them. “I remember being horrified by this story,” Lublin said. “ ‘You never even met her! How could you know that you’d hire her or not? Just based on what she looked like?’ And he’d say, ‘Yes, and that’s why you need to go comb your hair.’ ” Lublin chose to respond with a kind of benevolent pragmatism: in 1996, at the age of twenty-four, and with the help of three nuns in Spanish Harlem, she founded Dress for Success, a nonprofit that provides interview suits to underprivileged women looking for jobs. The budget came from five thousand dollars that she had recently inherited from her great-grandfather and her occasional winnings at poker. The nuns found space in a church, rented for a few hundred dollars a month, but it flooded a week before the launch, and Lublin was forced to transplant all the garments to her apartment, a one-bedroom in Greenwich Village. Dress for Success’s inaugural client was Charline Brundidge, who had been granted clemency after a conviction for fatally shooting her physically abusive husband. As reported in the Times, “Gov. George E. Pataki gave Mrs. Brundidge a pardon. Dress for Success gave her a suit.”

Within two years, Dress for Success had expanded nationally and was operating out of thirty cities across the country. (There are now affiliates in sixteen countries worldwide, and the Home Shopping Network produces clothing for the organization.) But by 2002 Lublin was bored. She cashed in the bonds given to her for her bat mitzvah and left Dress for Success. She spent a year writing and fielding calls from headhunters intent on recruiting her for C.E.O. positions at large nonprofits. She had just turned thirty and knew that the companies approaching her saw her as a token young person. Then she got a call from the actor Andrew Shue, of “Melrose Place” fame, who had co-founded DoSomething.org ten years earlier. The organization, seriously in debt, had just lost its headquarters and almost all its staff. Lublin thought this would be the perfect place to start up again from scratch. This was 2003, a year before Facebook was launched, and Lublin knew that if the organization was to have a future it would have to live online. She closed five of DoSomething’s offices; reconfigured its board of directors; and began polling its teen-age members about their habits, preferences, and passions. The organization now operates on a healthy budget of more than nine million dollars, attracts corporate sponsorship from companies like JetBlue and H & M, and hosts benefits that raise up to a million dollars.

In the fall of 2011, Lublin began raising funds for Crisis Text Line. To date, she has raised about five million dollars. Promotion is solely by word of mouth, and within four months the organization was receiving texts from all two hundred and ninety-five area codes in the United States. Lublin, who is friends with many Silicon Valley entrepreneurs and sees herself as an iconoclast, has built Crisis Text Line more along the lines of a tech company than a nonprofit. She told me, “We think of ourselves a lot more like Airbnb or Uber or Lyft.”

Like a tech company, C.T.L. analyzes feedback from users, performs A/B testing, and is quick to make changes on the basis of what it finds. Although other data-driven philanthropic missions exist—Kiva, the microfinance site, and the public-school donation service Donors Choose are among the more well known—nonprofits have generally been reluctant to embrace methods of quantification that big corporations increasingly take for granted. But at C.T.L. the chief data scientist, Bob Filbin, was Lublin’s second hire. He co-wrote the data algorithms for C.T.L.’s system after travelling to crisis centers across the country and interviewing hundreds of volunteers about how their work could be made more effective. The communication techniques employed by C.T.L. counsellors are largely modelled on standard crisis-counselling practices, but C.T.L. has made modifications based on its data. It turns out that, for instance, statements couched in the first person (“I’m worried about how upset you seem”) are associated with positive responses.

The organization’s quantified approach, based on five million texts, has already produced a unique collection of mental-health data. C.T.L. has found that depression peaks at 8 p.m., anxiety at 11 p.m., self-harm at 4 a.m., and substance abuse at 5 a.m. The organization is working on predictive analysis, which would allow counsellors to determine with a high degree of accuracy whether a texter from a particular area, writing in at a particular time, using particular words, was, say, high on methamphetamine or the victim of sex trafficking. A texter who uses the word “Mormon” tends to be reaching out about L.G.B.T.Q. issues.

Out of consideration for texters’ anonymity, Crisis Text Line displays its findings only by state. (Arkansas ranks highest for eating disorders, Vermont for depression; suicidal thoughts are most common in Montana and least common in New Hampshire.) But eventually there will be enough data to allow the organization to confidently reveal Zip codes and area codes without the risk of making any single texter identifiable. Such a wealth of data is new in the field of mental health. Isaac Kohane, a pediatrician who also has a Ph.D. in computer science and is the co-director of the Center for Biomedical Informatics at Harvard Medical School, told me, “You cannot have accountable care—financially or morally accountable care—if you cannot count, and until recently we literally could not count with any degree of acceptable accuracy.” He added, “It’s been mind-boggling, to those of us who knew what was available, that Amazon and Netflix were creating a far more customized, data-driven, evidence-based experience for their consumers than medicine has.”

Lublin hopes that the data will eventually be useful to school districts and police departments. “The corpus of data has the volume, velocity, and variety to really draw meaningful conclusions,” she told me. Lublin also mentioned that many people have told her that she is “crazy” for not wanting to sell the data that have been collected. A hedge-fund manager said that he would happily pay for a subscription that allowed him access to crisis trends. “I was basically, like, You’re a jerk,” she recalled thinking.

One Sunday evening in early December, I embarked on a training course as a Crisis Text Line counsellor. I was out of the country, but, as befits an organization in which the person who saves your life may be thousands of miles away, training takes place online. You learn by watching videos, reading PDFs, taking online quizzes, role-playing with fellow-counsellors, and observing conversations live. Volunteers, who must be older than eighteen, have to pass a preliminary interview and a background check. The course runs for thirty-four hours, over a period of seven weeks, and concludes with a final one-on-one video interview lasting twenty-five minutes. New counsellors commit to one four-hour weekend or evening shift per week for a year.

At the start of the first session, the faces of twenty-one participants, on Webcam, appeared along the top of my screen, some small and framed by the rooms around them, others so close that I could see their pores. We smiled and waved at one another; we all looked to be in our twenties or thirties. Our supervisor, who was wearing a pink plush hat, introduced herself and, in a crackling voice, let us know that we were welcome to wear pajamas next time. Like the sessions that followed in the coming weeks, this one was ninety minutes long but felt more like thirty. The training combines two pleasures largely lacking in adult life: structured incremental learning and make-believe. In an instant-message box, we practiced replying to various imaginary texts. The recommended formula for replies is tentafier + feeling adjective + source of feeling (“It sounds like you’re feeling ashamed because your friend didn’t invite you to her party”). We also practiced paraphrased reflections (“You must be really upset with your friend”). We learned to ask open-ended questions and to actively identify a texter’s strengths: pointing out his bravery in reaching out, complimenting his self-awareness. Then we were paired up in order to practice these skills in a role-play, one of us pretending to be an upset teen-ager and the other acting as a C.T.L. volunteer. Afterward, we annotated the transcript with “pluses” and “wishes,” the organization’s preferred language for “good” and “bad.” Intermittently, the supervisor’s cursor appeared in the document to offer advice.

The thing I found most difficult was employing Crisis Text Line’s teachings while still coming across like myself. The maxim “Don’t sound like a robot” is often repeated, and eventually it was possible to achieve this effect by imagining my words being read by a teen-ager. Using as many contractions as possible came to seem surprisingly important, because formality gets in the way of affirmation. It was hard to fend off vague and echoey therapist-speak, and I wasted a lot of time trying to rephrase the question “And how does that make you feel?” before realizing that I didn’t have to. There is something humbling about Crisis Text Line, and, indeed, about help lines in general: a person in pain will say what she wants to say, and it probably doesn’t matter much who does the asking.

The weekly practice sessions are the core of the training. Volunteers also participate in two Observation Shifts (each three hours long), in which they have the opportunity to see actual conversations occur between texters and a counsellor. Crisis Text Line goes to great lengths to insure that texters’ identity remains secret, and trainees sign a stringent agreement to protect confidentiality. (I agreed not to divulge personal details.) During my first shift, I witnessed a halting conversation between a counsellor and a young girl with body dysmorphia. The conversation lasted for an hour and a half. The counsellor provided links to resources for people struggling with eating disorders, and the girl eventually agreed to distract herself with a bath and a movie. Simultaneously, that counsellor was texting with a girl who wanted to cut herself and was having suicidal thoughts, and with a third texter whose grades were plummeting because of depression. This last texter was much less engaged in the process than the others were. “This isn’t making me feel much better,” he or she wrote. Soon after, the communication fizzled out entirely.

A week later, I shadowed another counsellor. Her first conversation was with a girl who was fighting with her cousin and struggling against the urge to hurt herself. The next was with a college-aged young man who was confused about the romantic feelings he harbored for his ex-boyfriend, who had sexually assaulted him. The counsellor’s last conversation of the night was with the daughter of an abusive father. She wrote that she avoided him by spending lots of time locked in her bedroom. The counsellor reassured her and asked about her plans for the rest of the evening. She said that she was going shopping with her family, and that afterward she’d be alone. She typed, “Thats the part im scared for.” ♦