- Examine counselor and teen text exchanges to help counselors more effectively serve teens in crisis.
- Help counselors understand what factors could improve positive outcomes.
- Recognize patterns to allow Crisis Text Line volunteers to respond faster to potentially harmful situations and interact more effectively with teens in crisis.
- Uncovered a data deficit based on a low number of completed counselor surveys.
- Increased the counselor survey response rate from 20% to 75%.
- Helped Crisis Text Line understand what they needed to do going forward to better collect and use data.
Crisis Text Line’s mission is to become as widely known as 9-1-1 for young people in crisis. Teenagers love to text, pounding out an average of 3,339 texts per month. They also struggle with family problems, bullying, substance abuse and much more. Born from the nonprofit DoSomething.org and first described in a TED talk, Crisis Text Line turns the medium teens trust most, texting, into a way to help them meet life’s challenges, providing the support teens need during their toughest times.
Crisis Text Line is a free, 24/7 text line available nationwide that connects teens in crisis to trained specialists. Since their launch in 2013, CTL has exchanged over 4 million messages with teens in crisis, representing 87,000 conversations. Trained specialists provide advice and resources by text with teens in crisis. To date, Crisis Text Line specialists have counselled teens through nearly 2 million problems.
With 11,000 incoming text messages per day and an average exchange lasting 66 minutes, Crisis Text Line collects an enormous amount of anonymous data. To help specialists do a better job, Crisis Text Line turned to its huge database of interactions with teens, along with machine readable survey data from specialists ranking how well the text exchanges went. They wanted to examine a number of factors related to the specialist/teen exchanges to determine: How can past data help specialists improve positive outcomes? What can Crisis Text Line glean from past exchanges to improve policy and better help teens? Can data give specialists the power to predict what’s going on with teens and better respond in real time?
In November 2013, a team of DataKind volunteers, led by Data Ambassadors Arun Ahuja and Cathy O’Neil, partnered with Bob Filbin, data scientist at Crisis Text Line, during a weekend DataDive to sift through 6,400 conversations - a quarter-million messages - to determine the factors involved in a good text exchange. To protect the highly confidential nature of the exchanges, the volunteers used the specialist reports from each interaction, not actual message data. This included information like what time of day brought the most texts and how long the teens texted.
However, as the DataKind volunteer team began trying to answer the question at hand, they discovered a significant data deficit: specialists had filled out just 20 percent of the surveys, too small a sample to draw conclusions. Instead of calling it quits, the team decided to use the data they had to understand why certain reports were incomplete and if Crisis Text Line could improve the ways they collected data for the future. The volunteers analyzed the form for the best ways to make it easy for specialists to complete while still sharing the most meaningful lessons from their interactions with the teens. Based on this feedback, Crisis Test Line condensed the survey, and, within just three months, improved their response rate from 20 to 75 percent!.
Crisis Text Line has already begun implementing different tools for specialists based on their work with the DataKind volunteers. One such tool in the works will help specialists identify patterns based on words the teens use to help them ask more effective questions. Ultimately, asking better questions will help specialists more quickly help teens potentially in danger of imminent harm.
In addition, based on the success of a weekend-long DataDive effort, Crisis Text Line was able to collect more robust data on these interactions and has now launched a longer-term DataCorps project with DataKind to more quickly route teens to help based on their needs. Data on the volume, variety and velocity of teen crises could lead to smarter research, policy changes and more effective community efforts to support teens before they are in crisis.
DataKind Blog - Real World Change, 3 Months Later