Human, Model and Machine: A Complementary Approach to Big Data

Abstract

In this paper, we describe a framework for processing big data that maximizing the efficiency of human data scientists by having them primarily operate over information that is best structured to human processing demands. We accomplish this through the use of cognitive models as an intermediary between machine learning algorithms and human data scientists. The ACT-R cognitive architecture is a computational implementation of a unified theory of cognition. ACT-R cognitive models can take weakly structured data and learn to filter information and make accurate inferences orders of magnitude faster than machine learning, and then present these well-structured inferences to human data scientists. The role for human data scientists is both oversight and feedback; one complementary piece of a hierarchy of cognitive and machine learning techniques that are computationally appropriate for their level of information complexity.