Computer learning and risk assessment in child protection

Abstract

The volume and complexity of information in child protection cases means that there can be an overwhelming number of factors which seem pertinent to decision-making but which obscure any pattern within it. This paper examines the applicability of a technique known as computer learning to the area of risk assessment in order to extract any underlying patterns. The paper proposes first that there are a few key interrelated, broad-level concepts used to assess and thereby classify risk. These can be used as the basis for producing a set of rules under which a social work team operates. The classification of risk made by one social work team on 20 child protection cases was analysed to find underlying patterns of their decision-making. These patterns are presented in the form of ‘decision trees’, as a way of illustrating the group's past experience in assessing risk. The results are evaluated in terms of the complexity and plausibility of the decision tree produced. © 1998 John Wiley & Sons, Ltd.