018: How Crisis Text Line is saving lives with text messages and data

CrisisTextLine.org is a text message hotline that connects teens with counselors at any time of the day. They are now opening up their data to help professionals and politicians understand more about teen crisis. To date they have handled over 3 million texts and are now opening up this information for people to explore. We talk with Bob Filbin, the chief data scientist at CTL about how their system works and why it matters to open up their data.


Example of Crisis Data


Musical Thanks

Caroline Reese & the Drifting Fifth are a Pennsylvania-based alt-country/Americana band. They’ve performed throughout the East Coast, appeared on Gene Shay’s famous WXPN Folk Show, and recently opened for Ben Folds.


Episode 18
Speaker 1: This is Using The Whole Whale, a podcast that brings you stories of data and technology in the nonprofit world. This George Weiner, your host, and the Chief Whaler of wholewhale.com. Thank you for joining us. Have you ever been in a situation where you needed to make a personal call, but you were maybe in a meeting or around people you didn’t feel comfortable talking in front of? This was an issue that crisis text line identified as something that was stopping teens from reaching out to hotlines about the crises that they were dealing with. Welcome to episode 18, where we’re gonna be talking with Bob Filbin, the chief data scientist at Crisis Text Line, about how they’re trying to leverage the call data, the text messaging data that’s been coming in for Crisis Text Line and what they hope to do with it, to improve the lives of hundreds of teens in the united states.

Speaker 1: So I’m here in the Crisis Text Line offices, with none other but Bob. So, who are you Bob? What do you do here?{1;31}

Speaker 2: So my name’s Bob Filbin. I’m the Chief Data Scientist at Crisis Text Line.

Speaker 1: How long have you been with Crisis Text Line?{1:38}

Speaker 2: So I started actually before the system even launched, last August. I started with one other guy, the head of engineering, Chris Johnson, last February. And our goal was to build an app, a web-based app that counselors could log into and use that to text with teens in crisis.

Speaker 1: The concept sounds good, but tell me a little bit about some of your early impact and stats behind your year data analysis. You probably can’t wait. I’ll stop interrupting, you can tell me right now.{2:15}

Speaker 2: So we’ve already, in our first year, exchanged 3 million messages with teens in crisis.

Speaker 1: Geez.

Speaker 2: Yeah, which is amazing. The demand actually exceeds our supply in some ways, the number of counselors we have. So our big challenge right now is not getting the word out to teens in crisis. They’re using the service. It’s can we get enough counselors to respond to that demand.

Speaker 1: Interesting. So how are you finding these teens, or how, should I say, are they finding you?{2:40}

Speaker 2: So one of the most common ways teens that are finding us is through Google. So when a teen is feeling depressed, or in a moment even of thoughts of suicide, one of the things they often do is turn to Google, to search for information. Because a lot of times, these teens feel isolated, like they don’t have someone to talk to, and so they look to gather information in a safe way. And so when they see Crisis Text Line, they realize that they can actually interact with someone in an anonymous way, and get the information and support they need to move out of crisis. We have very generous support. We actually have $40,000 a month in support from Google Adwords to get the word out. We’re trying to use all of it, we’re not quite there yet, but it’s a great way to connect with teens.

Speaker 1: Alright, so I want to tease out a couple things here. Look, hotlines already existed. Like, I can call 1-800 I need help type of hotlines out there. I’m sure if I had Googled them I would have found it. Why is it that CTL, Crisis Text Line, is actually taking off in this way, you think?{3:42}

Speaker 2: So, Crisis Text Line is a critical move to a different medium that teens are using. There was an article in the Atlantic earlier this year that called texting the number one social media for teens. It’s the medium that they use and trust the most to communicate with friends. It also turns out to be, as a natural extension of that, as a great medium for them to connect with crisis counselors, often for the first time. So many of the teens who use our service are connecting to a crisis hotline for the first time. Some who have used other services before tell us, hey this is better than online chat or phone. And the reasons they give for that is like nobody can hear them texting. Oftentimes, crisis hit while these teens are in school. So maybe they go to a school bathroom. We’ve actually had teens text us from school bathroom stalls, where nobody can overhear them, nobody can see what they’re doing. So that privacy is a big part of it. One of the other things that they bring up is, I talked to a teen that said, I prefer texting to phone because I don’t want the counselor to hear me crying. And so it creates a safe space in the teen’s mind where they’re not as vulnerable as they might be on the phone.

Speaker 1: How does this service work? You mentioned we have more demand than we have supply. Talk me through how a text is handled and what you mean by supply and demand.{5:11}

Speaker 2: Yeah, so we have the teens in crisis who see some kind of marketing around, or maybe hear from a friend, about the service. And we have a short code, which is 741741. A teen will text 741741 and they’ll get an automated message back, saying, “Hey, thanks for texting us. What brings you here today?” Once the teen goes through that initial process, they’re connected with a counselor. The counselor is actually not on a phone, they’ll be sitting in front of a web-based app. So they’re sitting at a computer, just like you would do in G-chat, or Facebook messenger, and can send a message to the teen’s phone, and back and forth. So a conversation will last anywhere from 50 to 60 minutes, and the teen can text back any time in the future if they need further support.

Speaker 1: So 50 minutes is the average session. How many texts are you normally seeing in that session for that crisis?{6:10}

Speaker 2: It’s actually about 1 text per minute. And that can vary a lot though, so if, I text this way too, I’ll send a bunch of texts very quickly, and then maybe pause and talk to somebody else, send messages to somebody else. But there are spurts of texting, but on average it’s about 1 per minute. The interesting thing is texting is asynchronous for teens. Teens don’t expect ok, we’re in a conversation, so we’re gonna have a steady stream of texts going back and forth. Some teens will text for a minute, for ten minutes, just disappear for a few, and then come back. So our counselors are sometimes actually handling two conversations at the same time, because there are breaks in the conversation.

Speaker 1: So we’re averaging about 50 texts sent by the teen. Does that also merit 50 responses?{6:59}

Speaker 2: Yeah, so it’s actually about 50 or 60 messages total per conversation, so it’s about 25 or 30 exchanged each way. And it is pretty equal going from teen to counselor responding.

Speaker 1: Do you know from call centers, if that were an actual call, the average length of a hotline call?{7:19}

Speaker 2: The average hotline call if I remember, and we can double check this, but around 15 to 20 minutes. So it’s shorter, but actually the amount of time that our counselors and the texters and [unknown] is probably about the same, because there are those pauses and gaps in the communication.

Speaker 1: This is pretty incredible. It seems pretty simple on the surface, but talk to me about how you approached building this, because you were there from day one with Chris Johnson.

Speaker 2: Yeah.

Speaker 1: You’re sitting there being like, how the hell do we build safe, secure, and easy to use text messaging system for crisis?{7:58}

Speaker 2: So before going to that, one note on the last thread is I think the great thing about texting is that teens don’t interact in a constant way. For teens it’s not optimal for a moment of crisis. If a teen needs to duck out for five minutes and think through something on their own, or they’re out in a public place and they need to be interacting with other people, texting provides the flexibility to have that asynchronous conversation. And I think that’s optimal for a crisis, rather than being on the phone for 20 minutes consistently. The reasons for texting are clear, and we believe texting would be the right solution for teens. The adoption we’ve seen from teens is proof of that. So we came up with a strategy from the very beginning. We realized that as Crisis Text Line we’re a tech and data service at heart. We build the technology for counselors to use, and then we’re looking at the data to make sure our counselors are able to provide the best possible care for these teens. Our primary touch point actually as Crisis Text Line is building the technology for the counselors. They’re the ones going out and doing the direct service with the teens. So we built this app by me travelling actually around the country for a few months, interacting with over a dozen different crisis centers around the country, sitting down with counselors for a few days, watching how they’re using existing technology for online chat, or phone. And for a few small groups that are doing texting. And then getting the feedback from counselors and building that into the design of our app. So our app is really built by counselors and for counselors, so their feedback has been incorporated in. There’s really this human centered design approach, which I love. It’s about observing, it’s about interviewing and discussing. And little things like, when we first designed the app we had a welcome page. Okay, we wanna make this feel like a immersive and engaging experience for our counselors, make them feel loved and like a part of a team from the very first moment. But when I showed this to counselors, they said “why is there a welcome page? When I log into the system, the first thing I wanna do is start talking to a teen in crisis.” So get this welcome page out of the way basically.

Speaker 1: Stop welcoming me. Get me to what I need.

Speaker 2: Exactly. They’re just so mission driven, they want to get right to helping the teens in crisis, so we got rid of the welcome page.

Counselors are still limiting in factoring, so what we’re looking for is more people that are interested in volunteering. We actually recently launched a remote volunteer program. And so this is like a “Coursera” or a “udacity”, where you can take a course online. We built one for learning how to be an online counselor. It’s about 50 hours worth of training. So it’s an extensive process because it’s such critical work that these people are doing. But now you can do it all online, and from anywhere in the country. Where before, you had to be based in a city where Crisis Text Line existed to volunteer, which, there are many, but they tend to be in larger cities like Boston or Seattle, so suddenly we opened the doors to anybody who’s in Arkansas, or Texas, or some of these larger places that don’t have access to a local crisis center. So we’re hoping, and we’re seeing actually, that this is opening the floodgates for people who wanted to volunteer as a counselor for a crisis center and have never been able to before.

Speaker 1: So these volunteers, it seems like we need more of them. How many do you have, and how many do you need?{12:05}

Speaker 2: We have around 250 counselors right now who are volunteer counselors and we’re thinking about, okay, we’ve had 65,000 conversations in our first year. How do we move toward a rate of a million conversations a year? To do that, we think we would need around 3,000 counselors. So we’re looking to move from 250 to 3000.

Speaker 1: You recently did something which I think is surprising for a crisis text line, or even a crisis service. Which is, you opened your data, and this could be a scary turn, cause you’re like, wait a minute. There’s a lot of sensitive stuff being exchanged. So what have you exactly done, and why did you do it in regards to opening up your data?{12:45}

Speaker 2: We feel that our data is valuable in two ways. It’s a way for us to improve the service, to provide a better care to the teens who are using our service. The ultimate goal is always, putting teens number one, and providing the best possible care to teens in crisis. So we can look at how our conversations are going, and pass recommendations and trainings back to our counselors to say, Oh, it looks like perhaps, unsurprisingly, open ended questions are more effective than close ended question in engaging a teen in crisis. Insights like that are how we improve quality internally. We recognize though that there are practitioners, there are citizens, there are researchers, and policy makers out there who are trying to help teens in crisis but don’t have the data to know how to do that most effectively. With limited resources, limited dollars to spend against this, where should they put the money? So, for example, a national bullying association. What stage should they concentrate efforts on expanding their services in? Our data answers that question. For a local organization considering LGBTQ, what days of the week should that LGBT queue service be open for teens? It turns out that one of the most stressful days, one of the days teens struggle most with LGBTQ issues is on Sunday. So are these centers open on Sunday? Probably some of them are not. The open data focuses on this aggregate data, so it’s no personal information, no message level data, so you can’t actually read any conversations. This shows how these crises vary over time and by state. And so at a high level, organizations who wanna impact and improve outcomes for teens is crisis can see where they need to be directing their resources. It’s critical that we’re opening our data to help other organizations that really haven’t had the data around teens in crisis. I don’t know of another data set out there that’s offering real time, large look at how teens are experiencing crisis, by time, by state, and what issues they’re facing. And so empowering these organizations to spend their money wisely, to actually prevent these crises from happening. Hopefully that happens, and hopefully it does actually eventually put us out of business.

Speaker 1: So I know you’re a data guy, so I have to put you on the spot with this. You intend this to be used for up the river thinking by major organizations and institutions. Do you feel that you have a large enough database to handle insights coming from non urban areas?{15:22}

Speaker 2: Our data is definitely preliminary in that sense, we have about a year’s worth of data, 3 million messages, it’s a strong data set, but it will only grow in potency over time as more teens use the service, it will become more powerful. So we don’t have that level of precision yet, and I think what’s critical is making sure that this is part of a tool set that someone is using to help answer questions around how to help a teen in crisis. For example, looking at the top state for suicide, in our data set, where teens are thinking about suicide. Montana is the top state. If you look at the CTC’s data on where people are in a higher risk for suicide, it’s also Montana. So there is a lot of alignment between data sets. BUt it’s always important. There are differences when you get further down the list. It’s important to compare, what are the respected, well known resources and space about how teens are experiencing crisis. And we hope to be a part of that tool set.

Speaker 1: At what point do you hit that tipping point, we’ll say, where this becomes wildly useful, because we have the vine of data that we need.{16:35}

Speaker 2: Yeah, one of the things that we’re really excited about is, and one of the things that makes us different from a CTC or some of these other organizations collecting data is we’re moving quickly toward real time insights on how teens are experiencing crisis. So that’s the ultimate goal, and as we grow in size, we’re gonna move more towards. For example, last week with Robin Williams passing, we saw an immediate spike, 9 PM, traffic doubled, and consistently doubled over the next three days in comparison to our normal trend. So a very clear spike in trend from a public incident. And we want to be able to pass those insights on to the public immediately. We’re moving towards growing, and I think again, our greatest limitation is the number of volunteers. If we solve that problem, I think we’re going to get to that point where real time data is an incredible resource for people trying to help teens. We’re already there at state level, but precision will increase.

Speaker 1: So, let’s go back to you Bob. If I were creating a direct service organization, like a hotline or text line, to support teens, I don’t think I would put a data analyst on first. Well, I would, because I’m a goober geek. And I’m just curious, why do we need a data analyst on this team, as like part of the whole team? What is your role, and how has it shaped the direction of the organization?{18:03}

Speaker 2: One of the critical things that data can do for an organization is process information at scale. So when we’re talking about millions of messages, and tens of thousands of teens that we’re assisting. We have 65,000 conversations so far, we’re growing to repeat that, something like 450 conversations in a day, and so the scale is growing quickly. What are the insights that pop up that an individual looking that the information couldn’t see? So the goal for a data analyst is to turn these insights into processes for the organization. So can we find something like, oh for example, the number of messages that are exchanged in a conversation is highly correlated with the quality of that conversation. So more messages, higher quality. That’s an insight that wouldn’t come from the individual level. It comes from the data. And so being able to pass that insight back to the counselors is part of my role. It resonates with counselors when we talk about it. It’s like, oh yeah. What teens are really looking for in a conversation is someone to talk to. So the longer, the more engaged the conversation is, the richer the dialogue, and the more the teen probably derives from the conversation. So my role is turning information into processes, both externally, so towards these counsellors, but also internally, towards our staff. Thinking about how can we do what we’re already doing more efficiently? So we’re trying to recruit more counsellors, or get the word out, to teens in crisis. One thought on this is Google Adwords is actually a great tool, what’s the right message to send to make our service appealing and clearly valuable to a teen that’s going through depression or thoughts of self harm? And so we used data to shape the messaging, so I’m doing a little AB testing on what the most effective message is to show. And that increases the number of teens seeing our service, using our service, for the same amount of money. We’re investing the same amount of resources in Google Adwords, but now we’re doing it more efficiently.

What every organisation would benefit from data analyst is can we do the same thing that we’re already doing, but less resources spent on that. So less time and less money. And we’re constantly finding efficiencies that we can improve on the marketing side, but also on looking at how do we keep our counsellors feeling good about their experience? So we talked about “So we took away our welcome page,” well one thing that we looked into is what is the right color for our apps? So our logo is red, this bright bold red, it says Crisis Text Line. But then we realized though that might not be a good color for the app. It might, red can encourage feelings of anger or frustration, or just might be too intense for counselors who are sitting in front of a bright red screen for four hours. So what we did is we looked at phycological research, and it turns out there’s color, slight blue, that is the most calming, mind easing color out there. And so we actually implemented that, and saw a change on how counsellors were feeling about their experience in the app. Both anecdotally, and then through survey research.

Speaker 1: So talk to me. Open these data, and people can find it online, right. Is that officially launched?{21:59}

Speaker 2: It is officially launched, and yeah, people can find all the data online, and actually interact with the data, see I wanna look at this bullying by day of week, and what’s the hardest day. Or I want to look at LGBTQ issues by hour of day. So you can actually interact with our data to customize it to find the insights that you want.

Speaker 1: And that’s at CrisisTextLine.org?{22:20}

Speaker 2: CrisisTextLine.org/trends.

Speaker 1: As you were playing with this data, I’m curious, was there anything that surprised you as a finding?{22:30}

Speaker 2: One of the surprising things for me was around time of day. When our team is experiencing these crises, one simple insight was, okay, teens thinking about self harm. One spike happens when people would expect, sort of later on in the evening, 6 to 9 PM, when teens are maybe after school, after school activities, maybe eating dinner, maybe in their rooms working on homework, and alone. And so that’s a very intense time. The second highest spike though, which was pretty surprising to me, was 10AM. Teens thinking about self harm, when they’re in school. And so that’s something that I don’t think is intuitive, but revealed itself in the data. And there are spikes and stories like that for every single issue, where teens are experiencing these crises in ways that I didn’t expect, that experts in the space did not expect.

Speaker 1: As we get to move and wrap here, I mean we talked about our high level numbers of lets say 5000 conversations and 3 million texts. Are there any conversations that stay with you and it’s like, this is why I’m doing this.{23:40}

Speaker 2: We have the honor of being in the office with actually eight of our remote counselors. These remote counselors work for a parent company of ours, where the idea for Crisis Text Line came from, dosomething.org. And so eight counsellors at dosomething.org volunteer as crisis counsellors. So I’ve talked to one of those counsellors, Naomi, about her experience, and she said she was talking with a girl who was looking for a therapist, but she didn’t know anybody in her town who was a therapist, and she talked to a few people, but they weren’t able to help her. And she didn’t want to tell her parents. But Naomi was able to do some quick googling, ninja googling, and locate a local therapist for the senior youth. And the heartwarming outpouring from this girl to Naomi about basically changing this girl’s life and offering the consistent help that this girl needed. That’s amazing. Naomi loved that experience, and I loved hearing it from Naomi. How amazing this volunteer experience is for the people who are willing to commit the time to do it. It’s more incredible than any other volunteer community that I’ve done, and so it’s great to be able to hear it first hand.

Speaker 1: As we wrap up here, what do you guys need? Like what do you need to get to that next level at Crisis Text Line?{25:11}

Speaker 2: So one of the big things, I’m reiterating is volunteers. So people who are thinking about wanting to volunteer, this is a great opportunity. The people that go through it have life changing experiences, and changed lives. And so would love to have more and more people volunteering with us. We’re also thinking about strategies for growth. So thinking about how can we make this data useful to policy makers, researchers, citizens who want to improve situations for teens in crisis? So if you fall into any of those groups, or even any other crisis centers out there, we want to hear from you on how we can make this data useful to you. Because up top, on the site, we have this aggregate data that we’ve released to the public. We’re also working with a few workers at MIT Media Lab, and writing up at Stanford and Princeton on some researchers who are going to be doing some deeper dives into the data under very careful conditions, and deriving information from that to help more teens in crisis. And we want to help find people to partner with, and so we wanna hear from you on how this data can be useful to your mission.

Speaker 1: How do people find you on the Twitters or online?{26:40}

Speare 2: I am @BobFilbin, personally, and then for Crisis Text Line, @CrisisTextLine on Twitter and then crisistextline.org. If you wanna connect with me in particular, go to crisistextline.org/trends, and you’ll see all my favorite data, and you’ll see a box at the bottom to message me about the data.

Speaker 1: Alright, you heard it. Get in touch with this guy. They’re doing amazing work. Bob, thank you so much for your time.

Speaker 2: Thanks George.

Speaker 1: These are text messages that are really changing the lives of teens across the country, having tremendous impact on their mental health, and I really hope that the open data project does what it sets out to do, which is help inform the policy makers and organizations out there already working in the field to improve the lives of teenagers. I really hope that if you’re even slightly interested in this program, and potentially volunteering, that you’ll at least go to the site and check it out. I can’t imagine a better way to spend your time. That’s all we have for you today. As always, wholewhale.com/podcasts, and we have some helpful resources that can accompany this cast. As always, thanks for joining us. This has been “Using the Whole Whale.” For more resources on today;s show, please visit us at wholewhale.com/podcasts, and consider following us on Twitter, @wholewhale. And thanks for joining us.

Alright, I see you’re still listening, so might as well keep talking. This week’s music brought to you by Caroline Reese. Really fun take on country music, folk music kinda blended together. I urge you to check them out on bandcamp. I also have them listed on our site, wholewhale.com. Take care!