34 - Causes of Failure

Summary

INTRODUCTION

Often in life, things don't turn out the way we hope. Despite our best efforts to understand the world and plan a course of action that will achieve our goals, things go wrong somewhere along the way. When attempting to explain what went wrong, the explanations that people come up with are rarely very creative. It is much easier to rely on a certain pattern of explanation that is useful and satisfying across a wide range of failures than to understand really deeply the root of our problems. That is, it is easy to say that someone's plan didn't succeed because he or she didn't have the skills necessary to execute it properly or claim that the plan would have been successful if only there were a little more time, rather than deconstruct the myriad of causal influences that were factors leading to the failure to achieve a goal.

These patterns of explanation for failed plans received a significant amount of attention among Artificial Intelligence researchers in the 1990s, largely emanating from the Yale University / Northwestern University computer science labs led by Roger Schank. Schank (1986) elaborated the notion of an Explanation Pattern, and pushed a multiyear, multidissertation effort to author a cognitive model of casebased explanation built on this idea (Domeshek, 1992; Jones, 1992;Ram, 1989; Kass, 1990; Leake, 1990; Owens, 1990), summarized in (Schank, Kass, and Riesbeck, 1994). Within the context of this work, a significant amount of effort was put into authoring comprehensive taxonomies of the types of explanation patterns that people used in understanding failed plans. Primarily, elements in these taxonomies were used as indices for the retrieval of stored explanations in support of a case-based reasoning process. The aim was to develop an indexing vocabulary (or content theory) of plan failure explanation patterns with broad coverage and utility for both scientific and engineering goals.

In this chapter we develop a new taxonomy of plan failure explanation patterns by combining the best of the taxonomies that have been developed in the past, synthesizing a taxonomy that is more comprehensive and elegant than any of the individual taxonomies that have been proposed in the past, and integrate it into the overall framework of this book. We close the chapter with a discussion of the utility of our taxonomy in scientific and engineering research.