A New LLM-Based Paradigm for Offender Risk Assessment

Authors

  • Yintao Wang Zhejiang Police Vocational College, Hangzhou, China

DOI:

https://doi.org/10.54097/mzdke887

Keywords:

Large Language Models, Offender Risk Assessment, Paradigm Shift, Human-AI Collaboration.

Abstract

Traditional correctional risk‑assessment tools depend on static indicators alongside subjective human judgment, and they often fail to reflect real‑world conditions in both prison and community correction contexts. This work presents a different approach. Building on recent progress in large language models, we outline an alternative paradig centered on four capacities missing from existing tools: contextual understanding, causal reasoning, unstructured data processing, and dynamic learning. We further map these capacities to a practical system design, including three principles, a four‑layer architecture, and a closed‑loop workflow covering both prison and community correction. The resulting framework can serve as both a foundation for further research and a practical reference for system developers.

Downloads

Download data is not yet available.

References

[1] X. Q. Liu and A. Ma, "Research on intelligent offender risk assessment system based on big data," in Proc. 23rd Nat. Academic Conf. Psychology (Volume I), Chinese Psychological Society, 2021, p. 120, doi: 10.26914/c.cnkihy.2021.041892 (in Chinese).

[2] Q. J. Jiang and L. X. Lu, "Research on application scenarios of artificial intelligence and big data analytics in prison administration," China Prison Journal, vol. 39, no. 2, pp. 29–35, 2024 (in Chinese). DOI: https://doi.org/10.1002/9781394303601.ch3

[3] H. Lyu, "Research on offender risk assessment method based on big data technology," Studies in Crime and Rehabilitation, no. 8, pp. 45–49, 2020 (in Chinese).

[4] G. D. Gao, "Research on offender risk assessment method based on artificial intelligence," Studies in Crime and Rehabilitation, no. 3, p. 81, 2023 (in Chinese).

[5] Y. Zhou, Y. H. Ye and X. Y. Xie, "Thoughts on perfecting the risk assessment system for offenders in China," China Rule of Law, no. 9, pp. 106–112, 2024 (in Chinese).

[6] Z. T. Zhou, "Research on abnormal risk assessment method for specific populations based on time series prediction and text feature fusion," M.S. thesis, Shandong Univ., 2025, doi: 10.27272/d.cnki.gshdu.2025.005869 (in Chinese).

[7] J. L. Du, S. D. Chen, L. Ye, et al., "A survey of chain-of-thought technology in large language models," Radio Communications Technology, vol. 51, no. 5, pp. 877–887, 2025 (in Chinese).

[8] C. W. Ji, S. Y. Wang, T. S. Li, et al., "Interactive lie detection based on large models: Data and models," Journal of Chinese Information Processing, vol. 39, no. 5, pp. 129–139, 2025 (in Chinese).

[9] G. F. Ma, Z. X. Wang and S. L. Ma, "Risk prediction of prisoners based on big data," Journal of Hebei University (Natural Science Edition), vol. 36, no. 6, pp. 657–666, 2016 (in Chinese).

Downloads

Published

18-11-2026

Issue

Section

Articles

How to Cite

Wang, Y. (2026). A New LLM-Based Paradigm for Offender Risk Assessment. Academic Journal of Management and Social Sciences, 18(2), 66-69. https://doi.org/10.54097/mzdke887