Prototype Demonstration
DOC

AI Data Assessor

Rhode Island Department of Corrections

AI-driven classification assessment that conducts adaptive voice interviews across ten criminogenic domains, scores risk against four validated models simultaneously, and provides real-time analytical transparency throughout the process. Supports the full workflow from initial interview through board hearing decision.

Assessment Domains

Alcohol/Drug Use
Employment
Friends/Associates
Attitudes/Behaviors
Residential
Family
Aggression
Mental Health
Education
Criminal History

System Capabilities

Voice-Driven Adaptive Interviews: AI conducts structured interviews using text-to-speech and speech recognition. Questions adapt in real time based on responses, probing deeper when risk indicators surface.
Four-Model Risk Classification: Simultaneous scoring across Violent, Property, Drug, and General Felony risk models with empirically derived cut points and Criminally Diverse detection.
Ten-Domain Needs Assessment: Each domain scored 0-10 with contributing and protective factors. Mix of standardized static items and AI-generated dynamic follow-ups per SOW requirement.
Real-Time Analysis Transparency: Split-screen view shows AI reasoning, domain scores, and risk projections updating live during the interview. No black-box scoring.
Complete Classification Workflow: Interview through counselor review, record enhancement (enemy conflicts, parole dates, special needs), board hearing, and final classification with override tracking.
Compliance-Ready Architecture: Full audit trail, role-based access, containerized deployment. Designed for CJIS and HIPAA compliance mapping with INFACTS integration path.

How to Use This Demo

  1. 1Log in as a Counselor (Maria Rivera), Administrator (David Chen), or Board Member (Sarah Thompson) from the login screen.
  2. 2As a Counselor, select a subject from the pending queue and click "Start Interview" to launch the AI-driven voice assessment.
  3. 3Watch the real-time analysis panel as the AI interviews across all ten domains. Risk scores and reasoning update after each response.
  4. 4Review the completed assessment, add supplementary information (enemy conflicts, parole dates), and submit to the Classification Board.
  5. 5Switch to the Board Member view to review pending hearings, approve or override classifications, and see the full audit trail.
Powered by Claude · OpenAI TTS · Next.js · FastAPI