Friday November 12 14:05 – Friday November 12 14:40 in Planck Bohr

Synergizing model and human decision-making using augmented machine learning

Kay Hoogland

Prior knowledge:
No previous knowledge expected


Machine learning does not have to take over an entire business process to add value. When set up correctly, it can be of great help in supporting a business process. Our credit risk models are not fully in control of the decision-making. Instead, they provide predictions and explanations that lead to valuable insights for the people in our risk department.



For our credit risk modelling, we love to use machine learning models to support human decision-making. The past two years we worked hard towards making our models more transparent to support our risk department. We also democratised our data by making it available to everyone in the company. In my talk, I will introduce you to our approach of augmenting our machine learning and fuelling our feedback loop to continuously improve on our features and modelling approaches. Elements will include:

  • Co-designing a business rule model together with domain experts
  • Increasing transparency by sharing feature descriptions with our predictions
  • Gathering feedback by providing model explanations with Shapley values
  • How models and the risk department collaborate using challenges