Visualizing Machine Learning Uncertainty
Part of Project IKON
From qualitative research I conducted, we concluded that our use case called for an ML-driven visualization, as this could show stakeholders potential thematic overlaps between research projects and knowledge transfer activities. At the same time as we concluded that an ML-driven visualization was needed, we predicted that it would be challenging for the non-technical stakeholders to make sense of it. In this regard, I collaborated closely with my team colleague, Dr.-Ing. Christoph Kinkeldey, to conceptualize and design an ML-driven visualization and associated explanation techniques; and in particular, the visualization of uncertainty.
Kinkeldey, Christoph, Tim Korjakow, and Jesse Josua Benjamin. 2019. ‘Towards Supporting Interpretability of Clustering Results with Uncertainty Visualization’. TrustVis ‘19. The Eurographics Association. https://doi.org/10.2312/trvis.20191183.
Kinkeldey, Christoph, Claudia Müller-Birn, Tom Gülenman, Jesse Josua Benjamin, and Aaron Halfaker. 2019. ‘PreCall: A Visual Interface for Threshold Optimization in ML Model Selection’. In CHI ‘19 Workshop Human-Centered Machine Learning. Glasgow, UK. https://doi.org/10.17605/OSF.IO/XAZKT.