A preliminary report on the performance of CLASS.LD2

This study summarizes a portion of the validation data used in the development of CLASS.LD2, an expert system for classifying learning disabled students. Two hundred sixty-four students files were evaluated using CLASS.LD. There was disagreement between the multidisciplinary teams and CLASS.LD2 in 78 cases. These 78 cases were then submitted to three experts in LD classification. These experts were asked to make eligibility decisions for each of the 78 cases. Their decisions were compared to those of the expert system and the placement teams. The data from these comparisons suggested that (a) CLASS. LD2 was in agreement with the experts more often than were the placement teams, (b) the expert system's decisions were significantly correlated with those of the experts, and (c) in those cases in which the experts were in unanimous disagreement with CLASS.LD2, it was shown that CLASS.LD2 conformed strictly to state rules and regulations in making its eligibility decisions.