PW 1554 AI platform for responding to child maltreatment and simulating future recurrent risk

Background While there has been great progress in artificial intelligence (AI) and the Internet of Things (IoT), numerous social problems remain to be addressed. Child maltreatment is the one of the largest social issues in the world. Although some research projects have been using ‘big data’, few studies have used AI to analyze these data. In order to save children suffering from maltreatment, AI and big data analysis should be effectively employed. Purpose In order to effectively respond to child maltreatment cases, we compute the future recurrent risk and deploy our AI platform in the field, such as with local child protective services. Method and deployment In order to develop an AI platform, we used 3648 cases spanning 3.5 years in a prefecture in Japan as data, including demographic factors, risk assessment, and whether child protection services were involved. In addition, we built a tablet iOS app that enables local social workers to easily input and store the data into a cloud database. The platform can predict the future recurrent risk by Bayesian models for binary output once users can check the probability of the risk. Result The results of the present study revealed that future recurrent cases are highly associated with risk assessment factors, such as existing bruises and injuries on a child’s face, neck, or abdomen, whether children and/or parents want social workers to protect the children, parents’ aggressive attitude toward social workers, and child sexual abuse (including suspected) cases. Conclusion and future research We need to continue to collect data and implement the system in other local child guidance centers in Japan. Furthermore, the system should consider what type of support plan would be effective for decreasing future recurrent risk in order to provide a personalized decision-making support system for local practitioners.