Chapter, 2024

Towards Cognitively More Plausible Models

Series: Synthesis Lectures on Human Language Technologies ISSN 1947-4040, Volume 1653, Pages 153-158

Publisher: Springer Nature

DOI: 10.1007/978-3-031-43260-6_7

Contributors (2)

Beinborn L. (Corresponding author) [1] Hollenstein N. [2]

Affiliations

  1. [1] Vrije Universiteit Amsterdam
  2. [NORA names: Netherlands; Europe, EU; OECD]
  3. [2] University of Copenhagen
  4. [NORA names: KU University of Copenhagen; University; Denmark; Europe, EU; Nordic; OECD]

Abstract

When initiating the writing process for this book, we hoped that it would bring us closer to a definition of the properties that make a model cognitively plausible. Even when abstracting away from biological plausibility as we did in this book, we cannot yet identify a single approach that will simultaneously lead to higher cognitive plausibility on all of the three dimensions we discussed: behavioral patterns, representational structure, and procedural strategies. We think that it is more useful to interpret cognitive plausibility as a graded concept and evaluate models comparatively by clearly defining specific aspects of cognitive plausibility. Instead of focusing on models that achieve optimal performance under ideal conditions, we should emphasize aspects of impaired performance under degraded conditions [1].