Interviewing data - the art of interpretation in analytics

Abstract

Interviewing Data: The art of interpretation in analytics -- Algorithms and statistical models produce consistent results with confidence yet they do so with data that are subject to change. Furthermore, the underlying digital traces created within specifically designed platforms are rarely transparent. The emerging field which incorporates analytics, predictive behavior, big data, and data science, is still contesting its methodological boundaries. How can we use existing research tools to validate the reliability of the data? This paper explores alternatives to statistical validity by situating analytics as a form of naturalistic inquiry. A naturalistic research model, which has no assumption of an objective truth, places greater emphasis on logical reasoning and researcher reflectivity. "Interviewing data", based on journalistic practices, is introduced as a tool to convey the reliability of the data. The misleading 2013 flu prediction illustrates this approach and is discussed within the context of ethics and accountability in data science.