The Non-Optimality of Scientific Knowledge: Path Dependence, Lock-In, and The Local Minimum Trap

This arXiv paper argues that scientific knowledge at any historical moment represents a local optimum—constrained by cognitive biases, formal paradigms, and institutional lock-in—rather than a global optimum. Drawing an analogy to gradient descent in machine learning, the authors propose that science follows steepest gradients of empirical accessibility and institutional reward, potentially overlooking superior descriptions of nature. The paper identifies three mechanisms of lock-in and proposes concrete interventions for designing meta-scientific strategies to escape these constraints.