人工智能介入量刑机制研究——以认罪认罚案件为适用场域

Research on Artificial Intelligence Intervention in Sentencing Mechanism

认罪认罚从宽制度自试点以来,控审冲突、司法供给不足、量刑建议的规范化与精准化要求等层出叠现,亟待检察机关采取应对措施。检察机关将智能量刑机制引入认罪认罚量刑领域,逐步实现从传统公诉机关向智慧公诉机关转变,为解决上述问题提供了技术量刑路径。但司法实践显示,智能算法本身的技术性难题、可靠性难题、不周延性难题极大地阻碍了智能量刑机制的发展。为了更好地解决智能量刑机制在认罪认罚量刑领域的司法适用难题,可以采用理论预测与实证预测双轨并行互相检验的运行模式,保证量刑建议预测的精准性、规范性,并在此基础上加强对算法的理论设计,完善对当事人诉讼权利的保障。

Since the introduction of guilt and acceptance of punishment, problems such as conflicts in prosecution and trial, insufficient judicial supply, and standardization and precision requirements for sentencing recommendations have emerged one after another, and it is urgent for the prosecutors office to take corresponding countermeasures. By introducing the intelligent sentencing mechanism into the guilt and acceptance of punishment will gradually transform from a traditional public prosecution agency to a smart public prosecution agency, and at the same time provide a technical sentencing path for solving the above problems. However, combined with current judicial practice, it is found that the technical problems, acceptability problems, and perimeter problems have greatly hindered the development of intelligent sentencing mechanisms.In order to better solve these problems, we can adopt a dual-track operating mode of parallel and mutual verification of theoretical prediction and data prediction to ensure the accuracy and standardization of sentencing recommendations and predictions, and on this basis strengthen the theoretical design of algorithms and improve the protection of the rights.