Data-Driven Collaborative Decision Making in Complex Industrial Systems

Conventional AI models of supply chain management provide interesting recommendations, but often no real understanding and unreliable predictions, which reduces trust in their advice. In a research project with BMW, a team around Prof. Dr.-Ing. Hendro Wicaksono, Professor of Industrial Engineering at Constructor University, PhD-Student Ishansh Gupta and Master-Student Adriana Martinez has developed a pioneering method that closes existing gaps by taking a causal approach.

In addition to data relating to the reliability of suppliers, for example, human “tribal knowledge” from the company is also incorporated into the model. This makes its recommendations more accurate, trustworthy and realizable.

In a survey at BMW, all the experts questioned favored this causal AI approach over conventional machine learning models.  “The most critical and strategic business decisions are causal questions,” says Prof Wicaksono. “Causal models try to capture the process of data generation and not just fit curves to the data. They are not a black box but can answer the question of ‘why’.”