Identifying and predicting stereotype change in large language corpora: 72 groups, 115 years (1900-2015), and four...

Area of Science:

• Social Psychology
• Computational Linguistics
• Cultural Evolution

Background:

• Stereotypes are fundamental to social cognition but exhibit dynamic change over time.
• Understanding the stability and variability of stereotypes is crucial for social cohesion and equity.
• Previous research has lacked large-scale, longitudinal analyses of stereotype content and structure.

Purpose of the Study:

• To investigate the change and stability of manifest and latent stereotypes for diverse groups over 115 years.
• To examine the variability of stereotype change across different social groups.
• To identify linguistic and social correlates predicting stereotype change and stability.

Main Methods:

• Utilized natural language processing (word embeddings) on four large English-language text corpora spanning 115 years.
• Analyzed manifest stereotype content (top associated traits) and latent structure (trait embeddings, valence, warmth, competence).
• Quantified stereotype change, stability, and consistency across 72 diverse groups.

Main Results:

• Group stereotypes exhibited significant changes in manifest content but relative stability in latent structure.
• Substantial variability in change and stability was observed across the 72 studied groups.
• Linguistic factors (mention frequency) and social factors (group type, e.g., sociodemographic) predicted stereotype change.

Conclusions:

• A dissociation between manifest and latent stereotype change suggests enduring consequences despite shifting representations.
• Stereotype change is not uniform; specific group characteristics influence the rate and nature of historical shifts.
• Findings provide quantitative evidence for factors driving historical stereotype evolution, informing interventions and social understanding.