Ann Burr Winslow
DBA in Technology & Operations Management
Dissertation Chair: Prof. A. Edmondson
Using Learning Curves to Measure Performance Improvement: Models and Trade-Offs
Learning-curve research has found that rates of learning can vary across similar settings, such that cumulative experience is a necessary but insufficient predictor of learning-curve slope. In this research I consider situations in which a new technology presents opportunity for multiple dimension of performance improvement. The organization can learn to use the technology more proficiently. Alternatively, it could expand the domain of its application. I consider how attributes of the knowledge being learned (tacit versus codified) affect learning rates in different dimensions and the extent to which organizations face trade-offs between the two learning trajectories. Building on insights from studying learning curves in knowledge work settings, I then propose a new learning-curve model that includes parameters that, depending on research design, describe attributes of the knowledge being learned.
The analyses in Chapters 1 and 2 use data from 670 patients who underwent cardiac surgery performed by surgical teams using a substantially new technology. In Chapter 1, I find that improvement rates across organizations are more heterogeneous for dimensions of performance that rely on tacit knowledge than for those that rely on codified knowledge; that group membership stability predicts improvement rates for dimensions relying on tacit knowledge; and that when performance relies on codified knowledge, later adopters improve more quickly than earlier adopters. In Chapter 2, I find a tradeoff in rates of learning on two dimensions: improving proficiency with the technology (efficiency) and applying the technology to novel and more challenging uses (technical difficulty).
In Chapter 3, I develop a new learning-curve model based on organizational knowledge. I start with four assumptions about performance improvement: it is monotonic with knowledge, knowledge increases monotonically with experience, and learning is limited both by what is known and by what is left to know. From these assumptions I propose a differential equation as an alternative to existing learning-curve models. Its parameters relate to the organization's preparedness for learning and absorptive capacity for new knowledge, and also to the complexity of the knowledge being acquired and the extent to which it is codified and accessible.
