Most measures of residential segregation can be regarded as global measures because they are summary indices for the entire study region. But the level of segregation is not uniform within the region and therefore local measures are warranted. The entropy-based diversity index is a local measure, but it fails to account for interactions between population groups across areal boundaries. In this paper, the diversity index is modified to accommodate spatial interaction between population groups across areal units. In addition, a set of local segregation measures is introduced to model local segregation based upon the exposure concept in segregation studies and the spatial interaction concept in geographical modeling. A simulation and an empirical study are used to demonstrate the utilities of this modeling approach.