English
 
Help Privacy Policy Disclaimer
  Advanced SearchBrowse

Item

ITEM ACTIONSEXPORT
  Biases towards compositionally simpler hypotheses are robust and unaffected by learning

Rubino, V., Dayan, P., & Wu, C. (2023). Biases towards compositionally simpler hypotheses are robust and unaffected by learning. In 2023 Conference on Cognitive Computational Neuroscience (pp. 817-820). doi:10.32470/CCN.2023.1392-0.

Item is

Basic

show hide
Genre: Conference Paper

Files

show Files

Locators

show
hide
Description:
-
OA-Status:
Not specified

Creators

show
hide
 Creators:
Rubino, V, Author
Dayan, P1, Author                 
Wu, C, Author                 
Affiliations:
1Department of Computational Neuroscience, Max Planck Institute for Biological Cybernetics, Max Planck Society, ou_3017468              

Content

show
hide
Free keywords: -
 Abstract: Compositionality is an important and yet poorly understood feature of human behaviour. In this study, participants navigated mazes with hidden, compositional structure, which were generated using operations over spatial primitives. Although they were not informed about the underlying structure, participants improved their accuracy and decreased primitive-inconsistent actions over the course of the task. Participants also selectively tested hypothesis corresponding to compositionally simpler expectations (simplicity bias), with a large proportion of errors due to expecting greater compositional structure than present in the true path. However, this simplicity bias did not change over the course of the experiment, and remained robust throughout. These results suggest that the human bias towards simplicity is unaffected by experience with compositional structure, at least in the time-frame of our experiment.

Details

show
hide
Language(s):
 Dates: 2023-08
 Publication Status: Published online
 Pages: -
 Publishing info: -
 Table of Contents: -
 Rev. Type: -
 Identifiers: DOI: 10.32470/CCN.2023.1392-0
 Degree: -

Event

show
hide
Title: Conference on Cognitive Computational Neuroscience (CCN 2023)
Place of Event: Oxford, UK
Start-/End Date: 2023-08-24 - 2023-08-27

Legal Case

show

Project information

show

Source 1

show
hide
Title: 2023 Conference on Cognitive Computational Neuroscience
Source Genre: Proceedings
 Creator(s):
Affiliations:
Publ. Info: -
Pages: - Volume / Issue: - Sequence Number: P-2B.50 Start / End Page: 817 - 820 Identifier: -