When Perfect Targets Go Bad: The Pitfalls of Over-Optimizing Our Measures

I’ve always been a bit obsessed with how efficiency—something we usually consider an unalloyed good—can sometimes backfire. My interest in this topic was piqued by experiences in machine learning (ML) and Complexity research a long time back, where I realized that overfitting can make a model’s real-world performance worse even as its training performan…