Comparative Analysis of GPT Models for Detecting Cyberbullying in Social Media Platforms Threads

The escalating issue of cyberbullying on online social platforms has raised serious concerns regarding individuals’ mental well-being and emotional health. As Large Language Models (LLMs) gain popularity and prevalence, exploring their potential in detecting cyberbullying becomes crucial. This research conducts a comparative analysis between GPT-3.5 Turbo and Text-Davinci models to identify cyberbullying within Instagram conversation threads. The evaluation of accuracy, precision, recall, and F1 score against manually labeled data allows us to assess the performance and limitations of these models in the context of cyberbullying detection. Through our findings, we aim to shed light on the effectiveness of LLMs in addressing this pervasive issue and their impact on online safety and well-being.