关键词: 算法黑箱, 技术权力, 预测警务, 层级告知

Abstract:

Artificial intelligence crime prediction embeds itself into crime governance through a technical mechanism encompassing data input， algorithmic analysis， result output， and control measure implementation. Based on cognitive discipline， systemic architecture， and information asymmetry， crime prediction has evolved from an auxiliary tool into a form technological power， manifesting itself in the dominant power of predictive algorithms over the initiation of recourse， the transmissive power of prediction results in judicial proceedings， and the emergent technical autonomy generated by predictive systems. This engenders the risk of power expansion that erodes the principle of prosecutorial restraint， the exclusionary effect that weakens the participatory capacity of the accused， and the obfuscation of accountability chains that facilitates sanction evasion. A reasonable distinction should be made between prediction results and criminal prosecution. Differential evidentiary thresholds and presumption-of-harmlessness principles should be established. Tiered disclosure protocols and algorithmic interpretability standards should be implemented to ensure meaningful participation. Dual-review mechanisms with accountability frameworks should be instituted to strengthen substantive human intervention.

Key words: algorithmic black box, technological power, predictive policing, hierarchical notification