Thirty years have passed since Barry MacDonald published “Evaluation and the Control of Education,” in which he set out an ideal classification of evaluation studies in terms of their political assumptions. There were three categories, two of which were based on his knowledge of existing early practices. The first and most prevalent he called “Bureaucratic Evaluation,” defined as an unconditional service to government agencies that accepted their values and helped them to accomplish their policy objectives. The second he called “Autocratic Evaluation,” defined as a conditional service that offered external validation of policy in exchange for compliance with its recommendations.

But neither of these approaches, the first justified by the reality of power and the second by the responsibility of office, offered a solution to the ...