Contrast effects in typicality judgements: a hierarchical Bayesian approach

Area of Science:

• Cognitive Psychology
• Computational Modeling
• Concept Representation

Background:

• Investigates the internal structure of everyday concepts.
• Compares two computational models: Generalized Context Model (GCM) and Similarity-Dissimilarity Generalized Context Model (SD-GCM).
• Focuses on the role of contrast categories in typicality judgments.

Purpose of the Study:

• To examine the influence of contrast categories on the graded membership structure of everyday concepts.
• To determine if contrast effects exist and identify responsible categories.
• To assess if multiple contrast categories influence typicality.

Main Methods:

• Utilized computational models from artificial category learning.
• Employed a hierarchical Bayesian framework for model comparison.
• Analyzed typicality gradients for five animal and six artifact categories, considering all relevant contrast categories.

Main Results:

• Internal category structure is co-determined by dissimilarity towards potential contrast categories.
• In most instances, a single contrast category significantly contributed to typicality.
• Contrast effects appear to be more prevalent than previously assumed.

Conclusions:

• Dissimilarity to contrast categories plays a crucial role in shaping concept typicality.
• Findings highlight the widespread nature of contrast effects in everyday concept representation.
• Emphasizes the need to consider everyday concept characteristics when applying artificial category learning models.