Citation

Smith RB. Eval. Rev. 2002; 26(1): 3-39.

Affiliation

University of California, Santa Barbara, USA.

Copyright

(Copyright © 2002, SAGE Publishing)

DOI

unavailable

PMID

11840637

Abstract

The computerized fraud detector (CFRD) assigns suspicion scores to questionable automobile insurance claims. Evaluators pilot tested this algorithm in three offices, comparing its effects with three matched offices. Observers uncovered that in two target offices and one comparison office, Millennium 2000 (M2K) also was being installed. The study design thus became as follows: Two offices had two interventions, one office had CFRD but not M2K, another office had M2K but not CFRD, and two offices had neither. Hierarchical linear models document that offices with both new computer systems will have the most unfavorable employee attitudes toward computerized fraud detection, followed by offices with only one new system. Employees with jobs of higher rank and employees not receptive to innovation will dislike computerized fraud detection. Implementation of one computer system, CFRD or M2K, may have minor negative fixed effects on employee attitudes, but their effects on the between-office variance are inconsequential.

Language: en