Small Worlds and Big Data: Examining the Simplification Assumption in Cognitive Modeling

The simplification assumption of cognitive modeling proposes that to understand a given cognitive system, one should focus on the key aspects of the system and allow other sources of complexity to be treated as noise. The assumption grants much power to a modeller, permits a clear and concise exposition of a model’s operation, and allows the modeller to finesse the noisiness inherent in cognitive processes (e.g., McClelland, 2009; Shiffrin, 2010). The smallworld (or “toy model” approach) allows a model to operate in a simple and highly controlled artificial environment. By contrast, big-data approaches to cognition (e.g. Landauer & Dumais, 1997; Jones & Mewhort, 2007) propose that the structure of a noisy environment dictates the operation of a cognitive system. The idea is that complexity is power; hence, by ignoring complexity in the environment, important information about the nature of cognition is lost. Using models of semantic memory as a guide, we examine the plausibility, and the necessity, of the simplification assumption in light of big-data approaches to cognitive modeling.