Data Philosophy || Crisis Text Line
Crisis Text Line was built from the ground up around technology and data with the goal of helping people thrive.
We've collected one of the largest health data sets in the world. It's the only real-time data set of this size in the United States. And it is incredibly diverse--recognizing voices of various genders, ages, races and ethnicities, etc. Hooray for equitable representation in data sets! (This matters to us. It should matter to everyone!)
Our top goal is to support people in crisis. Everything we do with data drives this goal. We want to put more empathy in the world to help people live their best lives. To achieve this, we use data in two ways: (1) internally, to improve the quality of our service, and (2) externally, to improve the crisis space as a whole.
We find ways in which data can support our Crisis Counselors. They are on the front line, talking with thousands of texters in crisis every day. We use data to help them focus on doing what they do best: talking to texters in crisis. Here are a few ways.
+ Continuous Improvement of Training
Every day, our data reveals new patterns in what it means to provide effective crisis counseling by text. For example, we’ve found that Crisis Counselors who geniunely identify texter strengths (e.g., “you showed courage texting us.”) achieve higher satisfaction ratings from texters. Our data show that three most effective terms to use are brave, smart, and proud. (E.g., “That was brave of you to reach out to a friend.”) And in 2018, the word “impressive” surged as an effective word too.
+ Helping The Highest-Risk Texters First
Most crisis lines respond to texters in the order in which they arrive. We act more like a hospital emergency room, where a person with gun shot wound gets helped before a person with a broken leg. We call it texter triage. An algorithm runs in the background, and assesses a texter’s suicidal risk based on their first few messages. Texters at high risk get marked as “code orange” and move to #1 in the queue. During the Presidential Election, when we saw volume 8x our normal, we were still able to reach high risk texters in an average of 39 seconds. Shorter wait times mean lives saved. Read more about our innovative, ethical approach to queuing on our blog.
+ Texter Feedback
After every conversation, we ask texters if they want to share a message with their Crisis Counselor. These messages often confirm the impact the Crisis Counselor has had on a texter’s life. Here’s an example: “Kindness is magic and you’re a magician.”
+ Spike Detection
For crisis hotlines, demand has been unpredictable. For example, when Robin Williams died by suicide, many hotlines experienced a 2-4x increase in volume that lasted for days. For most hotlines, this can lead to long wait times for people in crisis. For people at risk of harming themselves or someone else, every minute matters. That’s why Crisis Text Line developed an algorithm that uses an ensemble of data sources to predict volume spikes, allowing us to spot spikes 6-8 faster than we could with people alone. This early warning means we can staff volunteers sooner. During the 5 spikes we’ve seen so far in 2019, we’ve served 93% of high risk texters in under 5 minutes.
We share data to support smarter research, policy, and community organizing. Unlike other large-scale datasets on mental health and crisis, our data has incredible volume, velocity, and variety. We’re using this in several ways:
+ Data Ethics Committee
Our Chief Data Scientist leads this team of preeminent health and science ethicists from academia and organizations. They advise us on data privacy, security, and ethics, and help us develop compelling research questions that help us put more empathy in the world and help people live their best lives. See our Committee Members.
Available to the public, this data updates automatically and shows insights like, which state experiences the most anxiety? Or what’s the worst time of day for substance abuse? In one case, our data showed that LGBTQ issues are lowest on Fridays. There is an LGBTQ organization that we spoke to which only offers texting services on Fridays. We recommended they switch their hours to Mondays! Check out the public data!
We work with partners ranging from the National Eating Disorder Association to the state of Montana. They use Crisis Text Line to support people in crisis in their communities, and we share aggregate, anonymous data back about the crises their communities are experiencing. (Interested in becoming a partner?)
+ Open Data Collaborations
We also share data with researchers through the Research Fellows Program.
The goals of the program are:
- Access. To open the door for world-class machine learning / artificial intelligence researchers to work with one of the country’s most important data sets.
- Dissemination. To create a consistent flow of innovative and insightful media stories to share with community groups and policy makers.
- Implementation. To create data science products that are implemented by Crisis Text Line and community groups, and improve the crisis space as a whole.
Our current projects:
- Researchers from the Lawrence Berkeley National Laboratory are attempting to predict and prevent military veteran suicides using deep learning and math. Read more.
- National suicide data is two years old. A team of researchers is using Crisis Text Line data, among other data sources, to forecast suicide rates, ulltimately leading to better prevention.
Explore learnings in the “Published Research” section below.
+ Published Research
Since 2013, 8 papers have been published based on our data, and another 8 are expected in the next twelve months. We’re proud to fuel this kind of learning! Here’s what has been published so far:
- Protecting User Privacy and Rights in Academic Data-Sharing Partnerships: Principles From a Pilot Program at Crisis Text Line; Anthony R Pisani (University of Rochester Medical Center) et al.; J Med Internet Res 2019.
- Finding Your Voice: The Linguistic Development of Mental Health Counselors; Justine Zhang (Cornell University), Robert Filbin, Christine Morrison, Jaclyn Weiser, Cristian Danescu-Niculescu-Mizil (Cornell University); Proceedings of ACL 2019.
- Crisis support-seeking behavior and temperature in the United States: Is there an association in young adults and adolescents?; Margaret M. Sugg (Appalachian State University), P. Grady Dixon (Fort Hays State University), Jennifer D. Runkle (North Carolina State University); Science of The Total Environment 2019.
- Child Maltreatment Disclosure to a Text Messaging–Based Crisis Service: Content Analysis; Laura Schwab-Reese (Purdue University), Nitya Kanuri (Yale University), Scottye Cash (The Ohio State University); JMIR Mhealth Uhealth 2019.
- Crisis Text Line use following the release of Netflix series 13 Reasons Why Season 1: Time-series analysis of help-seeking behavior in youth; Laura K. Thompson (Appalachian State University), Kurt Michael, Margaret M. Sugg; Prev Med Rep. 2019.
- Adolescents in crisis: A geographic exploration of help-seeking behavior using data from Crisis Text Line; Laura K. Thompson (Appalachian State University), Margaret M. Sugg, Jennifer R. Runkle (North Carolina State University), Social Science & Medicine 2018.
- Large-scale Analysis of Counseling Conversations: An Application of Natural Language Processing to Mental Health; Tim Althoff (Stanford University), Kevin Clark, Jure Leskovec; Association for Computational Linguistics 2016.
- Mixed-Initiative Real-Time Topic Modeling & Visualization for Crisis Counseling; Karthik Dinakar (MIT Media Lab), Jackie Chen, Henry Lieberman, Rosalind W. Picard; MIT Media Lab 2015.
Loris.ai is a for-profit company created with the purpose of bringing more empathy into the world. Crisis Text Line has the founder equity in Loris--meaning, we own a large chunk of the company ensuring that the love of people comes first. Team Empathy! Loris leverages insights from Crisis Text Line’s techniques and aggregated data learnings to develop enterprise software to help companies take better care of customers. The result: better customer service for humans and sustainable funding for Crisis Text Line (so we never have to throw a fundraising dinner!). We do all this while keeping texter privacy front-and-center; no Personally Identifiable Information is shared with Loris.
So, thats a good summary of how we think about data and what we’re doing with it. We love learning. Got ideas for how we can do a better job of championing humans and putting more empathy in the world? Have a research idea that will help the world better understand anxiety or self harm or something else? Want to work here? For all that and more, email email@example.com.