A Multi-Scenario Approach to Continuously Learn and Understand the Evolution of Norm Violations