Representation, Provenance, and Explanations in Database Theory and Logic (Dagstuhl Seminar 24032)

Abstract

This report documents the program and the outcomes of Dagstuhl Seminar "Representation, Provenance, and Explanations in Database Theory and Logic" (24032), which was broadly in the area of database theory. Database theory formalizes the theoretical underpinnings of databases and analyzes them with mathematical tools. We focused on questions related to the fundamental problem of efficient query evaluation: compute the answers of a query on a database. This seminar focused on three key aspects of query evaluations. (1) Representation studies the tradeoff between expressivity, compactness, and efficient computation of outputs from the inputs, including circuits and knowledge compilation forms, enumeration, and direct access. (2) Provenance captures the computation process of outputs from the inputs using a compact formula, and has applications to probabilistic databases. (3) Explanations give meaningful insights to responsibilities of different inputs toward an output beyond provenance, e.g., by using Shapley Values from co-operative game theory that has been recently popular in both DB and ML.