Can AI Sentiment Analysis Detect the Risk of Social Exclusion in SNS Conversations Before It Happens?

Early Detection of Social Exclusion Through Sentiment Analysis: Today, cyberbullying is growing rapidly at a startling rate, causing serious emotional and psychological harm to children and young adults. Social exclusion in online settings is recognized as one of the primary driving forces to this issue which is why it is essential to be able to identify these exclusionary signs early on is crucial to successful intervention. Thus, this study evaluates the potential of AI models like Open AI’s ChatGPT in detecting early signals that indicate social isolation in simulated social networking services (SNS). A dataset of 6 conversational scenarios (bullying, neutral, and positive) with a total of 24 utterances were created and assessed with sentiment labels, risk categories, and sentiment intensity scores. The evaluations from ChatGPT’s algorithm were then compared to human-coded annotations. Bullying scenarios consistently scored higher in the risk range (0.7-1.0) with strongly negative sentiment tables, whereas neutral and positive scenarios clustered near zero, demonstrating clear polarizations. Additionally, there was a high degree of agreement between human annotations and model evaluations especially when it came to distinguishing between neural encounters and high-risk bullying. Contrastingly, it failed to show an accuracy with identifying more subtle kinds of exclusion capturing the nuances behind expression. Results reflected that GPT’s model served to be an effective system for early warning toward dangerous interactions in social media settings. Although there is limited ecological validity due to the conversation data being simulated, the results act as a fundamental paradigm for using sentiment analysis to reduce cyberbullying.