From Notepad AI to Social Media: How Can Text Style Transformation Mitigate Social Harm?

Researchers from the University of Dhaka and other institutions have proposed a novel writing-assistance framework that reduces societal harm by transforming aggressive, toxic, or emotionally harmful social media comments into softer, more neutral stylistic forms. The approach, inspired by Notepad AI (a simple AI writing assistant), applies controlled stylistic modifications rather than censorship, preserving core informational content while lowering emotional intensity and identity-based attacks. The framework introduces the Emotion Drift Index (EDI), a metric that systematically quantifies emotional change before and after rewriting. This allows platforms to evaluate the effectiveness of stylistic rewriting in reducing harmful interactions.

The work addresses a gap in content moderation: most current efforts focus on detecting and removing harmful posts, but less attention has been given to mitigating harm through stylistic transformation while keeping meaning intact. By leveraging the neutral, minimalist tone of Notepad AI, the framework can transform a hostile comment into a constructive one—e.g., changing a personal attack into a factual critique. The EDI metric provides a measurable way to ensure emotional intensity drops sufficiently. The authors argue this approach could reduce online toxicity without suppressing free speech, offering a scalable alternative for social media platforms. The paper is available on arXiv (2604.27365) and pending conference submission.