Understanding the Practice of Discovery in Enterprise Big Data Science: An Agent-based Approach

Abstract

Scientific discovery is substantially a social process. It involves organizational and inter-personal dynamics, resource and data constraints, biases and fads, as well as serendipity and chance encounters that are usually hardly represented in formal depiction of discovery. In this era of big data science, with heavy reliance on crowd-sourced data, open innovation, and collaborative analytics, the effect of the social and material realms on the process and practice of discovery is likely to become more acute. Understanding, and possibly predicting, the roles of these new data practices, organizational dynamics, and social infrastructures in shaping discovery can inform the design of more effective tools for enterprise big data science. In this paper, we present an agent-based model of the practice of discovery in big data science. Using a simulation system based on the practice-based approach to work study, the concept of bounded rationality, and the Gaia methodology for simulating organizations, we model big data science as an activity occurring within the social and organizational context of an enterprise. We present the background of this work, give an overview of the conceptual design of the model, and show some initial results.