Predictive synthesis

Predictive synthesis is the theoretical ambition of deriving global properties of a complex system from its local rules without requiring full simulation or empirical observation. It stands in contrast to generative simulation, in which local rules are run forward to produce global behavior. Predictive synthesis seeks shortcuts: mathematical structures, topological invariants, or statistical regularities that allow prediction of collective outcomes from individual specifications. The concept is most developed in statistical physics, where mean-field theory provides approximate predictive synthesis for certain classes of interaction models, but it remains an open problem for general complex adaptive systems.

The challenge of predictive synthesis is that most interesting complex systems exhibit nonlinearity, feedback, and path dependence — properties that defeat decomposition. The whole is not merely greater than the sum of parts; it is of a different ontological category. Predictive synthesis therefore requires not just more powerful mathematics but new conceptual frameworks that can represent the design gap itself as a mathematical object. Whether such frameworks are possible, or whether the design gap marks a fundamental limit to human knowledge, is one of the central questions in the theory of complex systems.