Driving innovation and AI with advanced visualization

Fredrik Moeschlin, Volvo Group, argued that the hardest part of data science in a company of 100,000 employees is neither the models nor the data, but the communication with the experts who have to act on the results, such as a gearbox engineer. He used the classical triad of logos, ethos, and pathos: facts are not enough without trust in the analyst and some care in how the results are shown.

He built the presentation itself in Processing, with vehicle movements on a 3D globe and data from OpenStreetMap in place of Volvo’s connected-vehicle data. He showed how plotting raw positions invites patterns that are not there, how feature importances from tree-based models narrow the search, and that three findings out of four, in his experience, turned out to be consequences of a fault rather than its cause. The work therefore has to be done together with the domain experts.

In the questions, he compared the trust that data science needs with the trust that passengers place in aviation.