EVALUATION CAN CONTRIBUTE TO POLICYMAKING AND PROGRAM ADMINISTRATION AND CAN PROVIDE EVIDENCE FOR PROGRAM ACCOUNTABILITY. AT THE OUTSET OF EVALUATION, IT IS NECESSARY TO KNOW THE PURPOSES TO BE SERVED AND TYPES OF DECISIONS AND POLICY QUESTIONS TO BE ENCOUNTERED; I.E., WHETHER PROGRAMS SHOULD BE EXPANDED, INSTITUTIONALIZED, MODIFIED, OR TERMINATED; RESOURCE ALLOCATION DECISIONS AMONG COMPONENTS; AND RESOURCE ALLOCATION DECISIONS ACROSS INSTITUTIONS AND AGENCIES. THE PRIMARY PURPOSE OF EVALUATION IS UTILITARIAN, TO PROVIDE EVIDENCE ABOUT PROGRAM EFFECTIVENESS SO THAT DECISIONS CAN BE MADE. PROBLEMS IN ATTEMPTING TO CONDUCT COMPREHENSIVE EVALUATION FOR BOTH ADMINISTRATORS AND STAFF RELATE TO MAJOR DECISIONS AND LARGE EXPENDITURES INVOLVED IN PROGRAMS AND PROGRAM EVALUATION, THE NEED FOR EXPLICITLY DEFINING AND FIRMLY CONTROLLING A PROGRAM, AND THE ISSUE CONCERNING THE SCALE OF EVALUATION (SCOPE AND COST). THE CLOSEST THAT RESEARCHERS HAVE COME TO IDEAL POLICY EVALUATION IS THROUGH SOCIAL EXPERIMENTS. SOCIAL EXPERIMENTS HELP PLANNING FOR THE FUTURE, BUT ARE COSTLY, AND THERE ARE FEW PLAUSIBLE NEW PROGRAMS TO TEST. THEREFORE, TO EVALUATE EXISTING PROGRAMS EVALUATORS MUST BE BE GIVEN CONDITIONS AND CONTROL TO DO THE BEST POSSIBLE RESEARCH; THIS MEANS SHIFTING AUTHORITY OVER CERTAIN PROGRAM DECISIONS TO THE EVALUATION TEAM. POLICY DECISIONS THAT COMPLETELY ELIMINATE ONGOING PROGRAMS ARE RARE. USE OF TRADITIONAL EVALUATION METHODS CAN BEST PROVIDE COMPARATIVE STUDIES OF THE EFFECTS OF DIFFERENT PROGRAM COMPONENTS AND/OR ON DIFFERENT POPULATIONS. TO AVOID PROBLEMS WITH EVALUATIONS PERFORMED ONLY ONCE, IT IS SUGGESTED THAT PROGRAM INFORMATION-GATHERING INCORPORATE EVALUATION DATA. COMPUTERS MAKE SUCH DATA COLLECTION POSSIBLE. EVALUATIONS SHOULD BE ADAPTED TO THE NEEDS OF INDIVIDUALS WHO WILL USE THE FINDINGS. NOTES ARE INCLUDED. (DEP)