Toward Digital Well-Being: Using Generative AI to Detect and Mitigate Bias in Social Networks

A machine learning pipeline is proposed to detect and mitigate bias in social network content to improve digital well-being. The system uses RoBERTa and DistilBERT models to classify content from Reddit and Twitter, identifying both explicit and implicit bias with high accuracy. Once bias is detected, a generative AI model (ChatGPT) crafts a context-sensitive response through a virtual moderator persona. This mitigation strategy adapts its tone, offering firm corrections for explicit bias and educational suggestions for implicit bias, aiming to foster a more constructive online environment.