Provenance for online decision making

Abstract

It is commonly believed that provenance can be utilised to form assessments about the quality, reliability or trustworthiness of data. Once presented with contradictory or questionable information, users can seek further validation by referring to its provenance. While there has been some effort to design principled methods to analyse provenance, the focus has mostly been on offline use of provenance. How to use provenance at runtime, i.e., as the application runs, to help users make decisions, has been barely investigated. In this paper, we propose a generic and application-independent approach to interpret provenance of data to make online decisions. We evaluate the system in CollabMap, an online crowd-sourcing mapping application, to make decisions about the quality of its data and to determine when the crowd’s contributions to a task are deemed to be complete.