The Ombudsman: A Pattern Language for Forecasters

Abstract

In recent decades, people have made numerous attempts to produce collections of principles for one art or another in areas outside management science. One well-regarded work is that of architects Christopher Alexander et al. (1977). In their classic book, A Pattern Language, they tried to identify and describe a comprehensive set of principles or patterns to guide the practice of architecture and urban design. Published in 1977 by Oxford University Press and containing over 1,000 pages, the book is still a best seller. An example of the sort of thing one learns there is Pattern 33 on night life: “Most of the city’s activities close down at night; those which stay open won’t do much for the night life of the city unless they are together (p. 180).” The idea of using such pattern languages has gained a following among computer-software designers. Coplien and Schmidt (1995), for example, present 30 papers from the first conference on pattern languages of program design. In his classic text on human-computer interaction, Schneiderman (1997, p. 74) identifies important “underlying principles of design,” such as these: enable frequent users to use shortcuts, offer informative feedback, and design dialogs to yield closure. In the following set of papers, forecasting experts examine the impacts of an effort to collect such principles for the science of forecasting. Forecasting is a complex task, one at which we have been notoriously bad over the years. Whether we are predicting which technologies will gain widespread acceptance, whether mergers will be successful, or how rapidly a disease will spread, we often miss by large amounts. Yet, in practice, the value of good forecasts is indisputable. So, the effort of Scott Armstrong and his 40 collaborators to articulate what we know about improving forecasts in terms of principles that can guide practice is important to business research. As Armstrong and Pagell (2003) note, while thousands of articles have been written about forecasting, managers must overcome formidable obstacles to make sense of this material. These problems are not exclusive to the forecasting enterprise. Management is besieged by a virtual explosion of knowledge, much of it highly conditional. In a recent lecture at the Weatherhead School, for example, Karl Weick noted that organization theory has changed from a belief that organizational behavior could be described by a complete set of principles to a field in which more and more variables come into play. Because researchers keep adding variables but seldom remove them, our models grow ever more complex. Rather than enhancing the confidence of those who must rely upon them, we reduce both their confidence and their ability to act. We clearly need new ways to organize our knowledge. The forecasting principles project has produced some 139 general principles and many additional specific or conditional ones. One view might be that this wealth of principles represents a pretty complex model. On the other hand, because the principles summarize and evaluate a massive literature and have been organized to reflect the structure of common forecasting tasks, they have the potential to simplify access to knowledge about forecasting. In a sense, many of the variables that have been proposed in the literature have been dropped.