Discussion of Dutta’s Paper

Summary At the beginning the author formulates the implementation problem (with complete information). Then he gives a short historical survey of implementation theory. Naturally, the classical works of Hurwicz (1960, 1972) and the seminal work of Maskin (1977) are mentioned as the cornerstone of implementation theory. The description of the present state of the theory relies on a recent survey by Moore (1992) on implementation with complete information; the key sentence from Moore is: ‘We suddenly seem capable of implementing anything, provided that we appeal to the right notion of equilibrium.’ However, Dutta points out that most game forms produced so far by the implementation exercise have undesirable properties that render them theoretically unacceptable and useless for applications. As a solution to the problem of the ‘irrelevance’ of implementation theory (his terminology) he suggests the ‘pragmatic approach’: in each problem of implementation we should concentrate on the specific structure of the set of outcomes (that may be, for example, a topological space or a measure space) and the special properties of the Social Choice Correspondence (for example, Pareto optimality, neutrality, anonymity, or restricted domain), and look for ‘simple’ and ‘natural’ game forms which implement the given Social Choice Correspondence.