Summary: | Signal detection
theory (SDT) was developed to analyze the behavior of a single judge but also
can be used to analyze decisions made by organizations or other social systems.
SDT quantifies the ability to distinguish between signal and noise by
separating accuracy of the detection system from response bias - the propensity
to over-warn (too many false positives) or under-warn (too many misses). We
apply SDT techniques to national and state-level data sets to analyze the
ability of the child welfare services systems to detect instances of child
maltreatment. Blacks have higher rates of referral and the system is less
accurate for them than for Whites or Hispanics. The incidence of false
positives - referrals leading to unsubstantiated findings - is higher for
Blacks than for other groups, as is the incidence of false negatives - children
for whom no referral was made but who are in fact neglected or abused. The rate
of true positives - children for whom a referral was made and for whom the
allegation was substantiated - is higher for Blacks. Values of d' (signal
strength) are roughly the same for Whites, Blacks, and Hispanics but there are
pronounced group differences in C (a measure of the location of the decision
threshold). Analyses show that the child welfare services system treats Blacks
differently from Hispanics and Whites in ways that cannot be justified readily
in terms of objective measures of group differences. This study illustrates the
potential for JDM techniques such as SDT to contribute to understanding of
system-level decision making processes.
|