Query-Answer Causality in Databases and Its Connections with Reverse Reasoning Tasks in Data and Knowledge Management

Causality is an important notion that appears at the foundations of many scientific disciplines, in the practice of technology, and also in our everyday life. Causality is crucial to understand and manage uncertainty in data, information, knowledge, and theories. In data management in particular, there is a need to represent, characterize and compute the causes that explain why certain query results are obtained or not, or why natural semantic conditions, such as integrity constraints, are not satisfied. The notion of query-answer causality in database was introduced in [86]. This notion is shown to be general enough to be applied to a broad class of database-related applications, such as explaining unexpected answers to a query result, diagnosing network malfunctions, data cleaning, hypothetical reasoning [86, 87, 84, 88]. In this thesis, we establish and investigate connections between query-answer causality and other important forms of reasoning that appear in data management and knowledge representation, e.g. consistency-based diagnoses [103], database repairs and consistent query answering [3], abductive diagnosis [35, 43], and the view-update problem [20, 77, 78]. These problems are classified in [83] as reverse data management problems. The unveiled relationships allow us to obtain new results for query-answer causality and also for the above mentioned related areas. Furthermore, we argue that causality in data management can be seen as a very fundamental concept, to which many other data management problems and notions are connected. In fact, we suggest causality as a unifying framework for reverse data management problems.