How AI helps volunteers support LGBTQ youth in crisis

Over 1.8 million LGBTQ youth seriously consider suicide in the U.S. each year. At The Trevor Project, an organization that provides crisis intervention and suicide prevention services to LGBTQ youth, the number of youth reaching out for support has ballooned since the start of the pandemic — at times nearly doubling our pre-COVID-19 volume. We knew we needed to train more volunteer crisis counselors to meet the growing demand for mental health support — and do so while working fully remote. The ability to connect the youth that we serve with highly trained counselors 24/7 is life-saving work, and will always hinge on human connection.

Over the last two years, has provided $2.7 million in funding and a team of nearly 30 Fellows to help scale The Trevor Project’s LGBTQ+ youth crisis support resources and technology using AI and machine learning. Most recently, Trevor and a team of Fellows built the Crisis Contact Simulator (CCS), a counselor training tool that uses AI to simulate conversations with LGBTQ youth in crisis. The simulator lets volunteer trainees practice realistic conversations with youth personas, equipping them with the skills needed to provide critical care. With this tool and other training innovations, we plan to grow our team of 700 digital volunteer crisis counselors by 10x! 

To become a volunteer crisis counselor, trainees learn about our counseling support model, active communication skills and LGBTQ identities, and take part in intensive one-on-one, human-led role play scenarios. We needed to build and test a tool that would provide a time-flexible, role-play opportunity for trainees outside of typical business hours — this was especially important since we know that nearly 70% of our digital crisis counselors volunteer on nights and weekends.  

To do so, we worked with the Fellows to bring together our knowledge and expertise in machine learning and natural language processing, product management, user experience, education, LGBTQ youth, and clinical psychology. “Through my work as a Fellow, I was able to adapt traditional user experience design practices and identify new ways for designers to collaborate with machine learning engineers. Being embedded with the Trevor project allowed us to be super collaborative with the training team as well, and gave us the opportunity to build lasting frameworks for their future AI work.” said Abby Beck, a UX Design Lead at Google. 

Thanks to six months of rigorous research, feedback, evaluations and data collected from thousands of role-play transcripts between the training team and volunteers, the CCS can emulate a number of digital youth personas. This allows trainees to practice realistic conversations with a wide range of life situations, risk levels and intersectional identities that span race, ethnicity, sexual orientation, gender identity, and more. “Riley,” the first digital youth persona, emulates a young person who’s struggling to come out as genderqueer. As more digital youth personas are introduced into the training program, Trevor counselors will become more adept at providing high-quality care and support to young people experiencing a variety of crises.