We are also indebted to numerous colleagues for their thoughtful criticisms and encouragement of an earlier draft of this paper, especially

No abstract

Estimating average causal effects under general interference, with application to a social network experiment. Annals of Applied Statistics, in press.

Before a social aggregate can be examined for its status as an organic system its status as entity must be evaluated. Indices of common-fate, similarity, and proximity may be appropriate to this task. Social groups as entities do not have an epistemological status Merent from such middle-sized entities as stones and rats, but are apt to be fuzzier, less discrete, less multiply confirmed, and in this sense less real. The degree of entitativity and the possibility of a sociology at a level of analysis separate from psychology is a matter for empirical determination rather than a priori decision.

HE United States and other modern nations should be ready for an experimental approach to social reform, an approach in which we try out new programs designed to cure specific social problems, in which we learn whether or not these programs are effective, and in which we retain, imitate, modify, or discard them on the basis of apparent effectiveness on the multiple imperfect criteria available. Our readiness for this stage is indicated by the inclusion of specific provisions for program evaluation in the first wave of the "Great Society" legislation, and by the current congressional proposals for establishing "social indicators" and socially relevant "data banks." So long have we had good intentions in this regard that many may feel we are already at this stage, that we already are continuing or discontinuing programs on the basis of assessed effectiveness. It is a theme of this article that this is not at all so, that most ameliorative programs end up with no interpretable evaluation (Etzioni, 1968;Hyman & Wright, 1967;Schwartz, 1961). We must look hard at the sources of this condition, and design ways of overcoming the difficulties. This article is a preliminary effort in this regard.Many of the difficulties lie in the intransigencies of the research setting and in the presence of recurrent seductive pitfalls of interpretation. The bulk of this article will be devoted to these problems. But the few available solutions turn out to depend upon correct administrative decisions in the initiation and execution of the program. These