VirtuGuard: Ethically Aligned Artificial Intelligence Framework for Cyberbullying Mitigation

Cyberbullying has become a concerning issue in contemporary society with the widespread use of digital communication tools and social media platforms. The impacts of cyberbullying can be far-reaching, especially for certain groups such as children and teenagers. This work aims to mitigate cyberbullying in an ethically appropriate manner with careful consideration of transparency, explainability, privacy protection, contextual understanding, and continuous monitoring and improvement. We propose an ethically aligned artificial intelligence framework for cyberbullying detection and analysis. The frame-work provides four core functions: 1) detecting cyberbullying comments with explanations; 2) building and enriching an evolutionary cyberbullying knowledge map with detected instances and external ethics resources; 3) constructing and maintaining a cyberbullying instance network; and 4) performing analytics and recommendation (e.g., mental aid support) based on the knowledge map and instance network. The output knowledge map, instance network, and analysis report collectively offer useful insights for policymakers, regulators, ethicists, industry stakeholders and researchers, facilitating the establishment of global standards and fostering collaborative efforts in addressing cyberbullying. We keep human in the loop and also ensure that user privacy is well protected.