Towards Interpretable Data Science

Explore the challenges and advancements in making data science more interpretable through this 35-minute lecture by Daniel Deutch from Tel Aviv University. Delve into the complexities of data processing for decision support, examining the issues arising from black box systems in data cleaning, database management, and machine learning. Learn about a holistic approach to interpretability focusing on counterfactual and attribution-based explanations. Discover key conceptual and computational challenges in the field, and gain insights into recent research findings. Examine relevant papers on explainability frameworks for data exploration, computing Shapley values in query answering, and constraints-based explanations of classifications to deepen your understanding of interpretable data science techniques.