CREATE IMPACT WITH TRUSTWORTHY AI - THE CASE OF ASSESSING BIODIVERSITY

GERARD SCHOUTEN

Biography
Gerard Schouten is professor at the Fontys University of Applied Sciences, school of ICT and leads the Fontys Knowledge Center Applied AI for Society. His research topic is AI and data. His interests include machine learning, deep learning, and in particular the creation of impact with trustworthy AI – explainable, fair and ‘green’ – for people and planet. Gerard is a valued member of the data advisory board of the province of Noord-Brabant. He also holds a position as guest researcher at the Naturalis Biodiversity Center where he focuses on applied AI for biodiversity.
Gerard graduated in physics, and has a PhD in the field of cognitive science (TU/e). He worked many years as a senior scientist for Philips Healthcare where he specialized in medical image processing and X-ray dose management. He has extensive experience in managing innovation projects and participated in many European research projects.

Abstract
In this presentation Gerard Schouten introduces a framework for creating societal and environmental impact with trustworthy AI. On top of human-centric quality aspects of trustworthy AI – such as model robustness, explainability, unfair bias, controllability, etc. – the framework consists of transdisciplinary process elements as a major condition for success. The framework is illustrated with a case that uses AI (computer vision) to gauge biodiversity, in particular wildflower richness and abundance. For this a unique expert-annotated reference dataset, with over 2500 images holding 150+ flowering plant species, is collected ‘in the wild’ (roadsides, urban areas, open fields) around the city of Eindhoven. The AI solution can be used by researchers and policy makers for large-scale automated wildflower monitoring and as an input to assess the value of so-called eco-systems services. The model can be embedded in a citizen science app to engage people for planet health.