Cyberbullying victimization identification and large language model-assisted assessment: a study of cyberbullying victimization lexicon construction and validation - Report - MDSpire
Advertisement
Cyberbullying victimization identification and large language model-assisted assessment: a study of cyberbullying victimization lexicon construction and validation
Clinical Report: Identification of Cyberbullying Victimization Using LLMs
Overview
This study developed a Chinese cyberbullying victimization lexicon utilizing large language models (LLMs) to enhance identification and assessment of cyberbullying.
Background
Cyberbullying is a significant mental health concern, particularly among adolescents, as it can lead to severe psychological consequences, including depression and suicidal behavior. Accurate identification of cyberbullying victimization is crucial for timely intervention, yet traditional methods face challenges due to self-reporting biases and help-seeking barriers.
Data Highlights
Dimension
Correlation (r)
p-value
Cyberbullying Methods
0.500
< 0.001
Perceived Harm
0.408
< 0.001
Coping Strategies
0.509
< 0.001
Overall Expression
0.870
< 0.001
Victimization Assessment
0.533
< 0.001
Key Findings
The lexicon consists of 442 words categorized into cyberbullying methods, perceived harm, and coping strategies.
Strong validity was established for identifying cyberbullying victimization expressions across all dimensions.
DeepSeek-R1 showed good performance in small-scale text classification but struggled with large-scale processing.
Significant discrepancies were observed between model outputs and human evaluations in vocabulary selection and weight assignment.
Human oversight is necessary for complex research tasks involving LLMs.
Clinical Implications
Reliance on LLMs should be complemented with human evaluation to ensure accuracy in large-scale applications.
Conclusion
The study emphasizes the need for human oversight in complex assessments.
