| Summary: | Individuals differ in their experience and expertise in identifying, categorizing, and recognizing real-world objects like birds, cars, and X-ray images. This dissertation sought to understand how and why they differ in these tasks by studying the underlying dynamics and representations of expertise through a series of experimental and modeling work, using bird expertise as an example domain. I first established a direct measure of bird expertise, to line up participants with varying levels of expertise along a continuum from novice to intermediate to expert. I then modeled the dynamics behind an expertise hallmark, the entry-level shift, using the sequential sampling models in a Bayesian Hierarchical framework. This mapped performance difference along the expertise continuum to difference in the underlying cognitive processes. Specifically, the difference in performances across the expertise continuum was largely due to the difference in the evidence accumulation rate, as well as a slight difference in the non-decisional processing time. The modeling results helped to differentiate two verbal theories of the entry-level shift phenomenon, specifically supporting the Differentiation Hypothesis but not the Basic-First Hypothesis. To further understand the representations that give rise to the difference in the cognitive processes, I then modeled the underlying representations of expertise by using the multidimensional scaling technique in a Bayesian Hierarchical framework, with the goal of mapping out representations along the expertise continuum and relating representations to performances. I found that participants along the expertise continuum shared the same perceptual space, but weighted the dimensions differently. I also found that participants with higher levels of expertise can better identify birds mainly because of their strategic weighting of the dimensions, but also because of their increased sensitivity to subtle difference between bird species. Together, these experimental and modeling work provided insights into the dynamics and representations of expertise, proposing a coherent computational understanding of real-world visual expertise.
|
|---|
