Professor Michihiro Kandori delivered a three-lecture series at HKUST on November 20–21, 2025, highlighting how AI and computational methods can advance economic theory. He showed (i) how large-scale data and machine learning improve behavioral models in strategic settings, (ii) how market prices can be understood as an efficient computational device solving complex allocation problems, and (iii) how AI-assisted numerical methods help tackle long-standing open questions in mechanism and matching design.

Key Points

• Machine learning uncovers richer behavioral patterns and outperforms traditional models in strategic experiments.

• Market mechanisms effectively approximate solutions to NP-hard allocation problems through decentralized price signals.

• Deep-learning tools (e.g., Rochet Net) help identify candidate optimal mechanisms in multi-dimensional screening problems.

• New results show when identical pricing across multiple items is revenue-optimal, even with heterogeneous distributions.