Evidence-Based Workplace Harassment Guidance System: Transforming #MeToo Narratives into Actionable Knowledge Through Machine Learning

Abstract Workplace harassment victims face a fundamental information asymmetry: organizations accumulate institutional knowledge about harassment cases, while individuals navigate these situations largely in isolation without systematic evidence about outcomes others experienced. This study presents an evidence-based guidance system that transforms 15,835 #MeToo narratives into personalized, actionable recommendations. The system employs dual sentiment analysis using BERT (98.1% accuracy) and VADER (73.4% accuracy), SimCSE embeddings for semantic similarity matching, and multi-label vulnerability detection across seven risk factors. Outcome analysis reveals sobering patterns: negative outcomes predominate (56–71% across vulnerability types), institutional inaction occurs in 16.9% of cases, and harasser accountability remains rare (3.6%). The system identifies 64 successful action sequences, with documentation-based approaches achieving substantially higher positive outcome rates. Professional evaluation with eleven practitioners (91% with harassment case experience) validates guidance quality: appropriateness M=4.09/5, usefulness M=4.09/5, actionability M=3.91/5, safety M=3.73/5, with 91% endorsement rate. External validation on 43 documented cases achieved 100% hit rate and 85.1% recall. This research demonstrates that machine learning can provide meaningful support for sensitive social domains when developed with rigorous professional validation and transparency about empirical realities.