A Framework for the Computational Linguistic Analysis of Dehumanization.

Area of Science:

• Computational Social Science
• Psycholinguistics
• Media Studies

Background:

• Dehumanization fuels intergroup bias, hate speech, and violence.
• Large-scale computational analysis of dehumanization is lacking.
• Understanding dehumanizing language is crucial for social justice.

Purpose of the Study:

• To develop a computational linguistic framework for analyzing dehumanization.
• To examine the evolution of dehumanizing language concerning LGBTQ people in media.
• To identify linguistic markers associated with dehumanization.

Main Methods:

• Created a computational linguistic framework based on social psychology research.
• Applied the framework to analyze New York Times articles (1986-2015) discussing LGBTQ individuals.
• Utilized natural language processing techniques to identify dehumanizing linguistic correlates.

Main Results:

• Found an overall trend towards more humanizing descriptions of LGBTQ people over time.
• Identified the label "homosexual" as more strongly associated with dehumanizing attitudes than "gay".
• Demonstrated linguistic variation and change in discourse surrounding marginalized groups.

Conclusions:

• The study provides a scalable method for analyzing dehumanizing language.
• Findings highlight the impact of specific linguistic labels on perceptions of marginalized groups.
• The framework has implications for detecting media bias and online abuse.