As artificial intelligence (AI) becomes a pivotal tool in decision-making, understanding its diverse impacts and designing systems that effectively complement human capabilities are essential. The symposium begins with an analysis of AI tools on volunteer-driven platforms like OpenStreetMap, highlighting how they boost productivity for new contributors but reduce engagement among experienced users. These trade-offs emphasize the need for strategically designed AI systems that balance efficiency with sustaining long-term community value. The next three papers then introduce approaches to improving and optimizing human-AI systems in practice: algorithmically targeting recommendations, strategically coarsening information, and providing actionable explanations. The first leverages a potential outcomes framework to model how recommendations influence human decision-making, proposing algorithms that specifically target areas where humans are prone to errors. This approach is demonstrated in hiring decisions, where managers evaluate candidates using both algorithmic assessments and private insights. The second focuses on calibrated coarsening, a method of adapting AI signals to address diverse cognitive biases and improve joint human-AI decision-making outcomes. This approach is applied in an experiment involving loan officers evaluating loan applications, showing that coarsening AI signals at optimal thresholds significantly improves decision accuracy. Ongoing work focuses on personalising and adapting this method to heterogeneous contexts and decision-makers. The third explores the use of algorithmic explanations, which foster trust and enable more effective reliance on AI recommendations by improving transparency and user understanding. This approach is explored in high-stakes contexts such as medical diagnoses and hiring decisions, where providing clear, actionable explanations of AI predictions can improve decision quality and trust in the system. Together, these studies offer a comprehensive roadmap for designing AI systems that enhance human capabilities, balance efficiency and engagement, and build trust in diverse, high-stakes contexts.
