JMIR Publications

ABSTRACT

Background:

Gelotophobia, the fear of being laughed at, is a social anxiety condition that causes discomfort and impacts daily life, self-esteem, and social relationships. Approximately 6% of neurotypical adults experience gelotophobia, but it affects 45% of individuals with autism spectrum disorder (ASD). Among adolescents with high-functioning autism spectrum disorder (hfASD), 41.98% exhibit moderate levels of gelotophobia, with higher rates observed in females. These statistics underscore the need for accurate detection methods, especially considering the social challenges associated with both ASD and gelotophobia.

Objective:

To develop a deep learning-based system that integrates facial emotion recognition and validated diagnostic tools to detect gelotophobia, particularly in individuals with ASD reliably.

Methods:

A model was developed to detect autism using deep learning algorithms. After training the model to identify autism, the data was fed into the pre-trained DeepFace model for facial emotion analysis to detect gelotophobia. This model was used to identify emotions such as ridicule or mockery in facial expressions, which may be associated with gelotophobia. For ambiguous cases, participants completed the GELOPH<15> questionnaire, a scientifically recognized tool for diagnosing gelotophobia. If the majority of questions were answered affirmatively, the system confirmed a gelotophobia diagnosis. The study measured prediction accuracy and validated the system's results across autistic and non-autistic groups.

Results:

A model was developed to detect autism using deep learning algorithms. After training the model to identify autism, the data was fed into the pre-trained DeepFace model for facial emotion analysis to detect gelotophobia. This model was used to identify emotions such as ridicule or mockery in facial expressions, which may be associated with gelotophobia. For ambiguous cases, participants completed the GELOPH<15> questionnaire, a scientifically recognized tool for diagnosing gelotophobia. If the majority of questions were answered affirmatively, the system confirmed a gelotophobia diagnosis. The study measured prediction accuracy and validated the system's results across autistic and non-autistic groups.

Conclusions:

This study demonstrates the effectiveness of combining deep learning techniques with validated diagnostic tools for the detection of gelotophobia, particularly in individuals with ASD. The high accuracy achieved by the system underscores its potential for clinical and research applications, facilitating improved understanding and management of gelotophobia in socially vulnerable groups. Further research could adapt this system for broader psychological assessments. Clinical Trial: This study is based on publicly available datasets (e.g., Kaggle) and does not involve direct medical interventions or randomized controlled trials (RCTs). Therefore, trial registration was not required. Ethical considerations were taken into account, ensuring that all utilized datasets were compliant with open-access policies and anonymized to protect privacy.