Large language models detect amplified abstract interpretation in trait anxiety

Area of Science:

• Cognitive Psychology
• Clinical Psychology
• Computational Linguistics

Background:

• Interpretation biases are linked to internalizing symptoms, but focus is mainly on negativity bias.
• Dual-coding theory suggests abstract and concrete processing influences cognition.

Purpose of the Study:

• To investigate how abstract vs. concrete elaboration in social interpretations relates to trait anxiety and depressive symptoms.
• To explore cognitive processing beyond emotional valence in internalizing symptoms.

Main Methods:

• 158 U.S. adults interpreted ambiguous social images under varying blur conditions.
• Large language models (LLMs) analyzed interpretations for emotional valence and abstract/concrete language.
• Trait anxiety and depressive symptoms were assessed.

Main Results:

• Both trait anxiety and depressive symptoms correlated with more negative interpretations, especially in negative scenarios (negativity bias).
• Individuals high in trait anxiety, but not depression, showed increased abstract language relative to concrete language across interpretation stages.
• LLM-derived linguistic indices captured these interpretation biases effectively.

Conclusions:

• Trait anxiety is associated with amplified abstraction in social interpretations, extending beyond negativity bias.
• LLM-based linguistic analysis offers a nuanced method for studying cognitive processing in psychopathology.
• Findings highlight the role of elaboration style in internalizing symptoms.