AI summary by TwinMind
The Trevor Project: Leveraging AI to Help LGBTQ Youth in Crisis | Talks at Google
Talks at Google
57m 54s
Introduction and Partnership Overview
- Jen Carter (Google.org tech/volunteering lead) hosts fireside chat with Trevor Project
- Trevor Project: world's largest suicide prevention/crisis intervention org for LGBTQ+ youth
- 1.8 million LGBTQ+ youth ages 13-24 seriously consider suicide annually in US
- Founded 1998 as first 24/7 national hotline for LGBTQ+ youth in crisis
- Serves 200,000+ young people annually via phone/chat/text
- Google.org + Googlers donated $3M+ over 2 years to advance Trevor's mission
- Sent 20+ Googlers as full-time pro bono fellows for 6 months each
Speakers and Roles
- Shira Kogan: Director of Corporate Partnerships at Trevor Project
- Works with corporate partners on social impact, DEI, employee engagement, cause marketing
- Helps companies contribute to mission of ending suicide for LGBTQ+ youth
- John Callerie: VP of Technology at Trevor Project
- Oversees product development, UX, AI, engineering, technology operations
- First technologist hired in org's 18-year history (4+ years at Trevor)
Trevor Project Program Structure
- Two main program buckets:
- Crisis counseling: free confidential 24/7 support for youth in crisis/suicidal ideation
- Prevention: building better world to prevent youth from entering crisis
- Creating safe, accepting environments for LGBTQ+ youth
- Empowering youth to live as authentic selves
Partnership Evolution and Components
- 2019: Applied to AI for Social Impact Challenge
- Selected as 1 of 20 winners from 2,600 applications
- Project focused on AI/ML for risk assessment to serve more LGBTQ+ youth in crisis
- 2020: Google reinvestment with additional funding
- Supported Trevor Space (global safe space social networking site)
- AI implementation for scalable volunteer crisis counselor training
- Human resources: 26 Google employees (full-time 6-month fellows + 20% program participants)
- External messaging/marketing collaboration
- Joint campaigns for Pride 2020 (adapted for racial justice context post-George Floyd)
- Google's reach helps Trevor connect with youth they couldn't reach otherwise
Technology Transformation at Trevor Project
Pre-Partnership State
- John was first tech hire in org's 18-year history
- Started with basic infrastructure needs: computers with tape holding panels, networking equipment
- Identity management and productivity software implementation
- Only couple hundred volunteers vs. 1.8M youth needing support
- Huge capability gap for scaling and improving quality of care
AI Impact Challenge and Early Development
- Built new customizable counselor platform
- Reached out to Google for creative solutions before formal partnership
- AI Impact Challenge application for problem traditional means couldn't solve
- Gained skills beyond technology: goal setting, OKRs, model development
- Google guidance on holistic approach: not just ML model but user experience integration
Current Technology Capabilities
- Full-time staff: product managers, UX researchers, UX designers, software engineers, ML engineers
- Team composition enables building "pretty much whatever we can think of"
- User experience research and usability testing standard practices
- Product development mentality organization-wide
Specific AI Projects Developed
Risk Assessment Tool
- Problem: Connect highest-risk youth with trained crisis counselors quickly
- Pre-chat intake form responses used to determine risk level
- Traditional approaches resulted in too many false positives
- First fellowship: 12 fellows in product, UX, software engineering, ML
- Breakthrough using NLP beyond classic classification models
- Used Google's ALBERT model successfully
- Has prioritized thousands of youth over past year
Crisis Contact Simulator
- Challenge: Scale counselor training from hundreds to thousands
- Traditional training: 4 role-play conversations with staff acting as youth personas
- 1.5 hours staff time per role-play session
- Solution: AI simulator for trainees to practice with
- Used captured conversations between training coordinators and trainees as input
- Generative NLP models to create conversations
- Launched January 2021 with first persona "Riley"
- Trained 700+ volunteers by July, over 1,000 by present
- Second persona "Drew" launching September 2021
Organizational Impact and Learning
Structural Changes
- Learned organizational frameworks from Google (OKRs now implemented org-wide)
- Tech team grew from 1 person to close to 30 people
- Understanding of bias in AI models and intersectional lens application
- Fairness considerations front and center in AI development
Recognition and Innovation
- MIT Technology Review coverage and other publications
- Recognized as innovative organization in technology space
- Better user experience for youth due to UX expertise
- Positioned as organization they "always wanted to be"
Knowledge Sharing
- Publishing articles with best practices for other nonprofits
- Innovation section on Trevor Project website with case studies
- Consultations with other nonprofits
- Encouraging nonprofits to take technology risks with willing funders/partners
COVID-19 Impact and Response
Operational Challenges
- Digital crisis counselors already remote, but phone lifeline had 2 physical call centers since 1998
- 2-week heroic effort to move call centers remote
- Purchased/mailed headsets to volunteers (noise filtering)
- Technical checks for voice over IP latency sensitivity
- Complete security model rebuild for off-site calling
Increased Demand and Reach
- Significant increase in crisis contacts (sometimes double pre-COVID volume)
- Huge increase in Trevor Space usage globally
- Youth cut off from supportive communities, sent back to unsupportive families
- Large proportion of LGBTQ+ youth live in non-affirming households
- Research: 70%+ LGBTQ+ youth reported poor mental health during pandemic
- Additional stressors: job loss, economic insecurity, discrimination (especially Black/Asian LGBTQ+ youth)
Positive Outcomes
- Now recruit lifeline volunteers nationwide vs. only West Hollywood/Manhattan
- Greater diversity in counselor base
- Fully remote organization with remote training capabilities
Future Plans and Scaling Goals
Growth Targets
- FY2021: Served 200,000+ crisis contacts (record high)
- Goal by end of 25th year (FY2023): 1.8M crisis contacts annually
- Confident in sustainable growth while maintaining/improving quality of care
Technology Development Pipeline
- Second crisis simulator persona "Drew" launching September 2021
- Cross-functional design vision sprint completed
- Reimagining counseling experience and youth experience
- Intentional AI applications in development
- Quality improvement through conversation assessment
- Real-time counselor support during conversations
- Resource recommendation systems
- Conversation summarization
- Trevor Space content flagging for crisis intervention
Global Expansion
- Trevor Space currently global
- Scoping language/international expansion for crisis services
- New team member leading global expansion work
Prevention Programs and Technology Applications
Program Areas
- Advocacy: ending conversion therapy state-by-state, locally
- Trevor Space: peer support social networking
- Research: annual LGBTQ+ youth mental health survey (34K-40K+ respondents)
- Education: training communities to support LGBTQ+ youth, create inclusive environments
Technology Applications in Prevention
- AI helping identify areas still supporting conversion therapy for advocacy focus
- Foundational UX research with 20 young people for Trevor Space improvements
- First Director of Trevor Space hired
- Technology foundation for scalable public training programs
988 National Suicide Prevention Initiative
- National suicide prevention number launching 2022
- Trevor Project working with government and nonprofits
- Bill passed unanimously - first to specifically call out specialized services for LGBTQ+ youth
- Recognition of need for more crisis organization resources including technology
Recommendations for Other Crisis Organizations
Challenges
- Tech industry moved faster than nonprofits, especially crisis-serving orgs
- Only couple crisis organizations have Trevor-level technical capabilities
- Need for more for-profit tech company philanthropic support
Solutions
- Crisis orgs should understand their value to potential tech partners
- Need trifecta: funding + pro bono support + technological expertise
- Significant time commitment from external technologists makes big difference
- Recognition that crisis organizations save lives doing critical work
- Suicide is 2nd leading cause of death among all young people
Google Volunteer Impact
- Google crisis counselors served 300+ young people directly
- Volunteered 250+ hours with Trevor Project
- Cross-functional Google support including Pride@ and Trans@ employee resource groups
