How to build Responsible AI System using RAI Dashboard in Azure ML

Artificial intelligence (AI) is rapidly transforming our world, from the way we work and communicate to the products and services we use every day. As AI continues to evolve and become more sophisticated, it brings with it enormous potential for positive change. This is where the concept of Responsible AI comes in. Although progress has been made in developing individual tools for specific aspects of Responsible AI, data scientists often need to employ a range of tools in tandem to gain a comprehensive evaluation of their models and data. If a data scientist uncovers a fairness issue using one tool, they may need to switch to a completely different tool to determine the underlying data or model factors causing the issue before taking any steps towards resolving it. There is no central repository to discover and learn about these tools, which prolongs the time required to research and learn new techniques. These different tools do not seamlessly communicate with one another. As a result, data scientists must grapple with wrangling datasets, models, and other metadata as they are transferred between tools. The metrics and visualizations offered by these tools are not readily comparable, and results are difficult to share. To address these challenges and make it easier to operationalize Responsible AI, Azure has introduced a Responsible AI dashboard (RAI Dashboard) which offers a unified view that allows for a holistic assessment and debugging of models, aiding data scientists in making informed business decisions.