1. ¹CESPRI, Università Bocconi, Milano, Italy
2. ²LEM – Scuola Superiore S. Anna, Pisa, Italy
3. ³Università G. D'Annunzio Chieti-Pescara, Italy
4. ⁴SPRU University of Sussex, Brighton, UK
5. ⁵Università dell'Aquila, L'Aquila, Italy

Correspondence: Luigi Marengo, LEM – Scuola Superiore S. Anna, Piazza Martiri Libertà, Pisa, I-56127, Italy. Tel: +39 05088343; Fax: +39 05088344; E-mail: l.marengo@sssup.it

Abstract

This paper discusses the issue of modularity from a problem-solving perspective. Modularity is in fact a decomposition heuristic, through which a complex problem is decomposed into independent or quasi-independent sub-problems. By means of a model of problem decomposition, this paper studies the trade-offs of modularity: on the one hand finer modules increase the speed of search, but on the other hand they usually determine lock-in into sub-optimal solutions. How to balance effectively this trade-off depends upon the problem environment and in particular on its complexity and volatility: we show that in stationary and complex environments there exists an evolutionary advantage to over-modularization, while in highly volatile environments, contrary to usual wisdom, modular search is inefficient in the long run. The empirical relevance of our findings is discussed especially with reference to the literature on systems integration.