Routines, Hierarchies of Problems, Procedural Behaviour: Some Evidence From Experiments

A laboratory experiment was performed as replication of the original one created by M. Cohen and P. Bacdayan at Michigan University. It consists in a twopersons card game played by a large number of pairs, whose actions are stored in a computer’s memory. In order to achieve the final goal each player must discover his sub-goals, and must coordinate his action with the partner’s one. The game therefore involves the division of knowledge and cooperation among players, and gives rise to the emergence of organizational routines. It is suggested that the organizational routines, i.e. the sequences of patterned actions which lead to the realization of the final goal, cannot be fully memorized because of their variety and number. It is shown that players do not possess all the knowledge needed by an hypothetical supervisor to play the best strategy: they generally explore only a limited part of the space of the potential rules, and therefore learn and memorize a simple, bounded set of "personal" meta-rules. These meta-rules, also called "production rules" in standard Cognitive Science’s language, are of the form . Each "Condition" can concern either the game configurations or the partner’s action. In the former case the identification of an appropriate "Action" depends on the sub-goals exploration. In the latter it depends on the recognition (or discovery) of interaction rules ; in this eventuality the production rule embodies a dynamic - and possibly cooperative - reaction to the partner’s action. Organizational procedures (routines) therefore emerge as the outcome of a distributed process generated by "personal" production rules . These routines, as in von Hayek’s view, "can be understood as if it were made according to a single plan, although nobody has planned it." (Hayek, 1980, p. 54). Empirical evidence is provided to support the above statements