Why are human animacy judgments continuous rather than categorical? A computational modeling approach.

Area of Science:

• Cognitive Science
• Computational Linguistics
• Psychology

Background:

• Animacy is often perceived as a simple binary concept (animate vs. inanimate).
• Human judgments of animacy reveal a more nuanced spectrum, including borderline cases and inconsistent classifications for certain entities and roles.

Purpose of the Study:

• To computationally model human animacy judgments.
• To identify features influencing animacy judgments using word embeddings and category information.

Main Methods:

• Utilized computational modeling to analyze animacy judgments.
• Employed bottom-up (word embedding components) and top-down (cosine distances to category names) predictors.
• Modeled both animacy and living/non-living judgments.

Main Results:

• Animacy judgments appear to stem from imperfect category membership estimates found in word embeddings.
• Cosine distance models successfully differentiated human (lower animacy) and animal (higher animacy) judgments.
• Results align with a family resemblance model for the concept of animacy.

Conclusions:

• Human animacy judgments are not strictly categorical but follow a family resemblance structure.
• Word embeddings capture nuanced semantic information relevant to animacy perception.