Explainable Detection

Explainable detection focuses on developing machine learning models that not only accurately identify events or patterns (e.g., AI-generated speech, traffic anomalies, medical conditions, online sexism) but also provide understandable justifications for their predictions. Current research emphasizes using model architectures like part-prototype neural networks and transformer-based models, along with explanation methods such as Shapley values and LIME, to achieve both high accuracy and interpretability. This field is crucial for building trust in AI systems and enabling responsible use in high-stakes applications like healthcare and online safety, where understanding the reasoning behind a model's decision is paramount.