人工智能輔助量刑場景下的程序正義反思與重塑

Reevaluation and Reshaping of Procedural Justice in the Scenario of AI-Assisted Sentencing

人工智能輔助量刑是我國近年來刑事司法人工智能應用的重大場景創新。在人工智能輔助量刑場景下，智能量刑輔助技術的適用不僅與實體結果相關聯，同時還與量刑程序深度融合，彰顯出兩種基本程序面向：在程序功能構造方面，其以規範量刑裁量權為程序功能取向，以量刑算法決策為程序功能實現路徑；在程序運行邏輯方面，則表現為“人機協同”型程序驅動模式以及“人主機輔”型程序責任分配格局。在此基礎上，智能量刑輔助技術的應用引發了有關量刑程序正義的三重結構性風險：即，理論層面的“傳統正義理論解釋力有限”、制度層面的“量刑算法決策正當程序機制闕如”、司法適用層面的“量刑裁判的人工智能算法依賴”。為保障人工智能輔助量刑場景下程序正義及量刑程序規範化的實現，應當提出具有針對性且體系化的風險治理方案。該方案具體包括：引入“以人為本”的技術性正當程序正義理論，強化理論供給；立足新型算法權利，建構訴訟化的量刑算法決策程序，填補制度空白；將智能量刑輔助技術的適用限定在輕罪案件場域、設置智能量刑輔助意見異議聽證程序，嚴格司法適用。

The AI - assisted sentencing is a significant scenario innovation in the application of artificial intelligence in recent years within China's criminal justice system. In the scenario of AI-assisted sentencing, the application of intelligent sentencing assistance technology is not only related to the result of substantial law but also deeply integrated with the sentencing process. It highlights two fundamental procedural aspects: in terms of procedural function construction, it focuses on the discretion of normative sentencing as the procedural function orientation, with the procedural function implementation path being the sentencing algorithm decision. In terms of procedural operation logic, it is manifested as a ''humanmachine collaborative'' procedure -driven model and a '' human lead - machine assistance'' procedure responsibility allocation pattern. Based on this, the application of intelligent sentencing assistance technology has raised concerns about three structural risks related to sentencing procedural justice: theoretical limitations in traditional justice theory, the lack of due process mechanism for sentencing algorithm decisions at the institutional level, and the reliance on artificial intelligence algorithms in sentencing judgments at the judicial application level. To ensure procedural justice and standardization of the sentencing process in the scenario of AI-assisted sentencing, targeted and systematic risk management solutions should be proposed. This includes: introducing a '' human-centered'' technical due process theory, strengthening theoretical supply; based on new algorithmic rights, constructing a litigation-based sentencing algorithm decision - making process, filling institutional gaps; limiting the application of intelligent sentencing assistance technology to minor criminal cases, establishing a hearing process for objections to intelligent sentencing assistance opinions, and ensuring strict judicial application.