Modeling the Other-race Advantage with PCA


The Other-race effect (ORE) refers to the well-known phenomenon of people being less accurate in recognizing faces of a different race. One popular hypothesis is that we learn to use face-features that are optimal for individuating faces of our own race; thus reducing the recognition accuracy for faces of a different race. However this hypothesis has not been able to explain some advantages other-race faces have in certain tasks. For example, some recent experiments showed that in a visual search task other-race faces are found faster than same race faces when the subjects show the ORE. A race based feature selection hypothesis where deviation from the familiar race is treated as an explicit part of the encoding has been proposed to explain this other-race advantage. In this paper, we argue that the other-race advantage can be explained without this assumption. We present the results from our simulations that suggest that the other-race advantage is an inherent characteristic of an optimal feature selection model.

3 Figures and Tables

Cite this paper

@inproceedings{Haque2005ModelingTO, title={Modeling the Other-race Advantage with PCA}, author={A. Haque and Garrison W. Cottrell}, year={2005} }