U.S. flag

An official website of the United States government, Department of Justice.

AI R&D to Support Community Supervision: Integrated Dynamic Risk Assessment for Community Supervision

Award Information

Award #
Funding Category
Competitive Discretionary
Congressional District
Past Project Period End Date
Funding First Awarded
Total funding (to date)

Description of original award (Fiscal Year 2019, $1,197,273)

RTI International and Applied Research Services Inc. (ARS), in collaboration with the Georgia Department of Community Supervision (DCS), propose to develop an artificial intelligence (AI) software tool termed IDRACS (Integrated Dynamic Risk Assessment for Community Supervision). IDRACS will incorporate dynamic risk factors and indicators to more effectively model the risk posed by individuals under community supervision and inform supervision strategies in real time.

RTI will employ a number of artificial intelligence (AI) classification models to predict recidivism, using model cross-validation to compare model type and specification to identify the most accurate models and reduce bias. To develop IDRACS, RTI will apply Machine Learning (ML) techniques to a historical dataset developed from case files of approximately 400,000 individuals on community supervision from 2016 through 2019.The data set will include criminal history and prison records, as well as data from DCS’s automated case management system. The proposed project will be completed in three phases over three years: Phase 1—creating the historical data set of individuals on community supervision and applying AI techniques to build and test models to identify differential risk; Phase 2— conducting two pilot tests — one to assess the value added by the IDRACS models and one to examine whether supervision practice data, including from body-worn cameras, will improve risk prediction; and Phase 3— developing an IDRACS dashboard and integrating it into the DCS’ case management system. IDRACS will be made available to NIJ for third-party evaluation. Deidentified data from model development will be archived.

Note: This project contains a research and/or development component, as defined in applicable law," and complies with Part 200 Uniform Requirements - 2 CFR 200.210(a)(14). ca/ncf

Date Created: September 12, 2019