Systematic iteration between model and methodology: A proposed approach to evaluating unintended consequences

This article argues that evaluators could better deal with unintended consequences if they improved their methods of systematically and methodically combining empirical data collection and model building over the life cycle of an evaluation. This process would be helpful because it can increase the timespan from when the need for a change in methodology is first suspected to the time when the new element of the methodology is operational. The article begins with an explanation of why logic models are so important in evaluation, and why the utility of models is limited if they are not continually revised based on empirical evaluation data. It sets the argument within the larger context of the value and limitations of models in the scientific enterprise. Following will be a discussion of various issues that are relevant to model development and revision. What is the relevance of complex system behavior for understanding predictable and unpredictable unintended consequences, and the methods needed to deal with them? How might understanding of unintended consequences be improved with an appreciation of generic patterns of change that are independent of any particular program or change effort? What are the social and organizational dynamics that make it rational and adaptive to design programs around single-outcome solutions to multi-dimensional problems? How does cognitive bias affect our ability to identify likely program outcomes? Why is it hard to discern change as a result of programs being embedded in multi-component, continually fluctuating, settings? The last part of the paper outlines a process for actualizing systematic iteration between model and methodology, and concludes with a set of research questions that speak to how the model/data process can be made efficient and effective.