How we use machine learning to protect young athletes while supporting their mental performance journey
Validated on 2,000+ diverse test cases including crisis phrases, sports metaphors, and normal athletic expression
Our Athlete Welfare Monitor uses a hybrid ML-enhanced triple-pass detection system designed specifically for the unique language patterns of young athletes. The system was developed with input from sport psychology research and validated against thousands of real-world scenarios.
140+ carefully curated crisis phrases are detected instantly with zero latency. This ensures immediate response for explicit statements of self-harm, suicide ideation, or farewell behaviors—no API delay when every second counts.
An advanced large language model analyzes ambiguous content with sports-context awareness. The system understands that "I killed it today" is positive athletic expression, not a crisis indicator. Age-appropriate prompting adjusts analysis for under-13, teen, and adult cohorts.
Advanced regex patterns catch complex multi-word constructions that might slip through simple keyword matching, adding another layer of protection for edge cases.
Immediate notification to designated safety contacts. Includes explicit statements of self-harm intent, suicide ideation, farewell behaviors, and passive ideation patterns.
Patterns indicating athletic burnout, identity struggles, social isolation, or chronic anxiety. These warrant coach attention but not crisis intervention.
Healthy athletic expression including sports metaphors ("killed it", "destroyed them"), normal frustrations, performance anxiety, and everyday challenges.
Inspired by FIFA player cards, ZenCard provides a visual representation of athlete mental resilience across three research-backed dimensions derived from the Martens/Vealey/Burton framework for competitive anxiety.
Confidence vs. Anxiety regarding sport performance. Tracks cognitive and somatic anxiety patterns.
Feelings of belonging vs. isolation. Measures connection with teammates and coaching staff.
External stressors including academics, relationships, and identity development outside of sport.
Athletes complete a 12-question baseline assessment to establish their personal mental performance profile. Questions are calibrated to age cohort.
Voice and text journal entries are analyzed by AI to adjust scores in real-time. Breakthrough patterns (anxious start, confident finish) positively impact resilience scores.
| Metric | Keyword-Only | ML-Enhanced | Improvement |
|---|---|---|---|
| Overall Accuracy | 76.56% | 99.5% | +22.94% |
| Crisis Detection | 93.79% | 100% | +6.21% |
| False Negatives (Missed Crises) | 37 | 0 | -37 |
| False Positives | 0 | 0 (In Test Set) | — |
Results from 2,000 validation test cases including crisis phrases, burnout indicators, sports metaphors, and normal athletic expression. Performance on future unseen data may vary.
ZenQuill combines the rigor of sport psychology research with modern AI safety practices to create a platform that truly understands young athletes. Our goal is to support their mental performance journey while maintaining the highest standards of care.
Learn More About ZenQuill