Responsible AI for Inclusive and Sustainable Workplace Communication: Detection of Misogynistic Language in Business Discourse

Abstract

Gender-biased language remains a significant yet frequently overlooked problem in corporate communication, with important implications for workplace equality, organizational trust, and sustainable human resource management. This paper applies Natural Language Processing (NLP) and deep learning to detect and address gender bias in business texts. Using a publicly available dataset of formal and informal corporate communication, the research integrates preprocessing, sentiment analysis, text feature extraction, and transformer-based classification to reveal implicit and explicit forms of discriminatory language. A fine-tuned RoBERTa model, enhanced with attention and dropout layers, achieves 91.2% accuracy, 88.7% precision, 90.5% recall, and an F1-score of 89.6%, indicating strong capacity to identify misogynistic content while maintaining a balanced relation between missed detections and false alarms. Beyond technical performance, the proposed approach is positioned as a responsible AI framework that supports inclusive workplace communication, ethical governance, and analytically informed diversity strategies. By connecting automated bias detection with organizational practices aligned with gender equality, decent work, and reduced inequalities, the study contributes to knowledge-based sustainable development. The findings indicate that computational language analysis can serve a broader organizational purpose by strengthening fair communication, accountability, and inclusive workplace cultures