Improving Human Decision-making by Discovering Efficient Strategies for Hierarchical Planning

Abstract To make good decisions in the real world, people need efficient planning strategies because their computational resources are limited. Knowing which would work best for different situations be very useful understanding and improving human decision-making. Our ability to compute those used limited small simple tasks. Here, we introduce a cognitively inspired reinforcement learning method that can overcome this limitation by exploiting hierarchical structure of behavior. We leverage it understand improve large complex sequential decision problems. decomposes problems into two sub-problems: setting goal how achieve it. discover optimal larger more tasks than was previously possible. The discovered better tradeoff between quality cost both existing algorithms. demonstrate teaching use significantly increases level resource-rationality require up eight steps ahead. By contrast, none previous approaches able performance on these These findings suggest our informed approach makes possible decision-making Future develop support systems world.