This paper studies an evolutionary programming technique, namely a genetic algorithm, to analyze how a population of decision-makers learn to coordinate the selection of an equilibrium or a social convention in a two-sided matching game with incomplete information. In the contexts of centralized and decentralized entry-level labor markets, evolution and adjustment paths of unraveling are explored using this technique in an environment inspired by the Kagel and Roth (2000. Quarterly Journal of Economics 115(1), 201-235) experimental study. As an interesting result, it is demonstrated that stability need not be required for the success of a matching mechanism under incomplete information in the long run.