Predictive insights through analogical reasoning: Application to screening new financial service concepts
Hoonyoung Lee · Scholarly Commons (University of Pennsylvania) · 1993
Analogical reasoning (AR) is a fundamental means by which managers learn from as well as solve problems based on their experience. However, the effectiveness of managers' analogical reasoning could be limited by the range of their personal experience, their limited cognitive capacities, and inherent biases. In this research, we propose to develop a computerized analogical reasoning system (ARS) that compiles individual managers' experience, and then selectively retrieves from it relevant analogies. Such pooling of experience is expected to enable managers to overcome some of the limitations associated with their (possibly) narrow range of experience. The system's automatic and controlled retrieval of past experience relevant to a present problem reduces the managers' cognitive effort involved in identifying a relevant analogy. Furthermore, the system could contribute to organizational learning as a computer-based organizational memory and information retrieval system, which--unlike human experts--can be easily accessed and shared. The dissertation is organized in terms of an introduction, three independent essays, and the conclusion. The first essay discusses the essence of our research by addressing questions such as what is analogical reasoning? why should we pay attention to it? and what are the problems associated with its application? It suggests an approach to overcome some of the problems. The second essay illustrates the suggested approach by developing a computerized analogical reasoning system (ARS) for assisting new product managers with screening new financial service concepts. The benefits of the system over other screening models and methods are discussed, and so are certain limitations. The third essay investigates into the forecasting ability of ARS using simulated data. It is demonstrated that ARS is more robust than regression models in handling forecasting problems of increasing complexity (i.e., increasing amount of error, fewer data points, greater number of variables, and increasing complexity of relationship between independent and dependent variables). The dissertation concludes by summarizing the contributions and findings. The conclusion also discusses a variety of other applications, the limitations of the current research, and directions for future research. The major contribution of this research is the development of an approach so that managers and organizations can optimize use of their past experience in learning and solving a great variety of business problems associated with developing and evaluating business activities, and forecasting.