What is the role of computational models in Cognitive Science? A quantitative and qualitative analysis of the history of the TRACE model of speech segmentation

Abstract

Implemented computational models are a central paradigm of Cognitive Science. How do cognitive scientists really use such models? We take the example of one of the most successful and influential cognitive models, TRACE (McClelland & Elman, 1986), and we map its impact on the field in terms of published, electronically available documents that cite the original TRACE paper over a period of 25 years since its publication. We draw conclusions about the general status of computational cognitive modelling and make critical suggestions regarding the nature of abstraction in computational modelling.

Vocal imitation can create acoustic attractors to guide mothers to pups in a crowded colony of Mexican free-tailed bats: A case-study of computational modelling in behavioural biology

“…Making the architecture of a computational model more elaborate is conventionally seen as contradicting the principle goal of modelling-simplicity. Chawla and Shillcock (2019) take a test case of a high-profile cognitive model and show that its citations overwhelmingly refer to the success of the model's architecture or to its algorithmic content, as opposed to elaboration of the model's architecture. There are, naturally, implemented models that have generated long-running, productive research programs, but even this outcome is not the same as the prospect of an ongoing exploration of open problems related to the original behaviour and leading back to the biological totality of the domain.…”