Determining predictions, actions, and consequences for Responsible AI

After the model is built and deployed in DataRobot, it might seem that our job is done—but not so fast. You should start analyzing what the predictions profile looks like and start discussing with users and stakeholders the details of actions to be taken. The models you have helped build are likely to introduce many changes in your system and will impact other people. It is therefore important to try to make sure that these impacts are not negative. Making sure that your models will not cause harm is called Responsible AI. This will build upon the work you did during the understanding phase through various diagrams.

Just as in previous sections we saw how a causal diagram helps you to relate features to a target, we can also see how the target affects other parts of the system. The diagram should reveal how the target impacts key objectives or outcomes; it should also reveal key feedback loops that will...