This project investigates the use of Natural Language Processing (NLP) and machine learning techniques to detect and explain social bias in short-form social media posts, specifically from platforms such as X (formerly Twitter) and LinkedIn. The study focuses on identifying gender, nationality, and profession-based bias through the development and evaluation of multiple detection systems, including rule-based methods, transformer-based neural models, and a hybrid architecture named PRISM (Probabilistic Reasoning for Interpretable Social Media Bias). The project also constructs and evaluates a benchmark dataset of annotated social media posts to compare system performance using standard NLP evaluation metrics such as precision, recall, macro F1-score, calibration, and confusion matrix analysis.