The aim: Develop novel AI and ML algorithms to guide decision-making to solve globally pressing problems accounting for uncertainty and limited resources.
So far, AI for decision-making has been developed mainly in the context of autonomous agents (closed loop), however CSIRO as a trusted advisor is often asked to provide recommendations for decision-making in human operated systems. This requires a shift of paradigm. The Decision activity developed algorithms that provided interpretable, explainable and trusted solutions for decision-making. This was achieved through a collaborative process of understanding and bridging the gap between application domains and AIML research.
Example of applications: adaptive decisions to protect our environment from invasive species, adaptive message framing to increase sustainable behaviours and optimal experimental design.
Keywords: Decision making and reinforcement learning, active annotation and Bayesian optimisation; verifiable, explainable, ethical ML/AI