PySS3: A Python package implementing a novel text classifier with visualization tools for Explainable AI

Explainability is built-in, not added later: Unlike many AI systems where explanations are retrofitted after development, SS3/PySS3 was designed from the ground up to visually explain its classification decisions, making it easier for users to understand why the system made a particular prediction.

Performance and transparency can coexist: The classifier achieved state-of-the-art results on risk detection tasks (CLEF's eRisk), demonstrating that explainable AI systems don't necessarily sacrifice accuracy—a key concern for organizations adopting responsible AI practices.

Open-source access enables broader accountability: By releasing PySS3 as an open-source package, the researchers made explainable AI tools publicly available rather than proprietary, allowing independent researchers and organizations to audit, test, and verify the system's behavior.

Real-world applicability matters for responsible deployment: The system was specifically designed to handle text streams and support incremental learning, addressing practical deployment scenarios (like monitoring for emerging risks) where AI systems must adapt and explain themselves in real-time rather than in controlled laboratory settings.

Visualization tools democratize AI understanding: The inclusion of visualization capabilities means that stakeholders without machine learning expertise—such as domain experts, ethicists, or affected communities—can potentially understand and scrutinize the model's reasoning.