Multidisciplinary characterization of embarrassment through behavioral and acoustic modeling

Area of Science:

• Psychology
• Affective Science
• Computational Linguistics

Background:

• Embarrassment is a common social emotion, yet it is under-researched compared to other emotions.
• It shares significant overlap with social anxiety (SA), a condition affecting a substantial portion of the population.
• Existing research has not fully elucidated the objective characteristics of embarrassment.

Purpose of the Study:

• To characterize the emotional state of embarrassment using an interdisciplinary approach.
• To develop and validate a behavioral paradigm for inducing and measuring embarrassment.
• To apply machine learning and acoustic analyses for objective emotion assessment.

Main Methods:

• A cohort of 33 participants described an embarrassing experience and read it aloud.
• Subjective embarrassment was measured via self-report scales.
• Objective analysis involved machine learning models for acoustic feature extraction and emotion classification (dimensional and categorical).

Main Results:

• Subjective embarrassment ratings increased after participants read their narratives aloud.
• Individuals with higher social anxiety scores reported greater embarrassment.
• Machine learning models achieved 86.4% accuracy in predicting embarrassment states; dimensional analysis (VAD) showed limited differentiation, while categorical analysis positioned embarrassment closer to boredom than sadness.

Conclusions:

• The study successfully characterized embarrassment through a novel behavioral paradigm and acoustic modeling.
• Objective acoustic markers for embarrassment were identified, showing potential as a biomarker for social anxiety.
• This research provides a foundation for further investigation into the objective measurement of social emotions.