| dc.description.abstract | Technology evolution prediction is critical for designers, R&D managers, and policy makers to make important design and R&D decisions and to develop effective government incentives. Many descriptive models (e.g., logistic S-curve model and Moore’s Law) have been developed for technology performance prediction, but these descriptive models do not identify what factors shape future technology performance and how designers, firms, or governments can manipulate them. In this dissertation, a quantitative ecological based theory is created for technology evolution prediction and manipulation. The quantitative ecological based theory consists of a Lotka-Volterra ecosystem model and a generic method for prediction intervals generation. The ecosystem model and the generic method are able to help designers, R&D mangers, and policy makers to predict technology technical performance (e.g., speed, capacity, and energy efficiency), to discern the causality of technology evolution, and to develop effective strategies to improve technology technical performance.
The Lotka-Volterra ecosystem model is extended from Lotka-Volterra equations in community ecology. The ecosystem model considers the interaction between a system technology and its component technologies in the relationships of symbiosis, commensalism, and amensalism. In addition, every parameter in the ecosystem model is associated with its causal factors, such as R&D investment and technical difficulty. The values and interpretations of these parameters are used to identify the key component technologies in a system technology and to develop effective strategies on improving system technology performance.
The generic method uses bootstrapping to generate prediction intervals for technology evolution. The prediction intervals help practitioners to assess future uncertainty and make contingency plans accordingly. The method can be applied to any prediction model based on mathematical functions or differential equations involving time. Parameter uncertainty and data uncertainty are considered in the method and the empirical probability distributions of these uncertainties are established. The appropriate confidence level α required to generate prediction intervals is determined using a holdout sample analysis rather than setting α=0.05 as is frequently done in previous research.
The quantitative ecological based theory is proven to be effective through four case studies of three representative technologies (i.e., concrete skyscraper, passenger aircraft, and central processing unit) in this dissertation. | en |
