AI LEGO: Framework for Cross-Functional Responsible AI Collaboration

A new study on arXiv introduces a concept called 'AI LEGO,' designed to improve teamwork in Responsible AI (RAI) practices during the early design stages. The researchers reviewed existing literature and spoke with eight industry experts to identify three main challenges: technical design choices often overlook non-technical team members; collaboration lacks a shared visual framework; and non-technical staff lack the necessary tools for assessing potential harm. Existing tools like JIRA and Google Docs aren’t quite effective. The proposed model features modular visual elements that bridge the gap between technical and non-technical team members, promoting a shared understanding and comprehensive harm assessment. It emphasizes the need to consider potential harms early in AI development from a socio-technical viewpoint.