洞悉黑箱:公众参与视角下众包算法审计机制研究OACHSSCD

Crowdsourced algorithm auditing,by mobilizing a distributed user group to observe algo-rithmic behavior and collect evidence,can effectively expand the coverage and problem identification capabilities of auditing.Through a"hierarchical-structural"dual-axis analysis framework,three typical models are identified.The platform-responsive model,where the platform leads the process,guided by compliance and efficiency,with structured but shallow public participation.The user-driven model,which emerges from the public's self-mobilization,highlighting grassroots power and external supervision capabilities,but with a relatively loose structure.The collaborative co-construction model,which integrates the platform,experts,and the public,seeking a balance between professionalism,transparency,and deep public participation in institutionalized collabora-tion.In the Chinese context,crowdsourced algorithm auditing is dominated by the platform-re-sponsive type,presenting three localized models:platform-integrated,third-party aggregated,and government-led.The advantage of the Chinese crowdsourced algorithm auditing model lies in its a-bility to quickly gather social feedback and trigger regulatory intervention.Public participation mainly plays a role through institutionalized complaint channels,and the public's role in agenda setting and in-depth auditing is still limited.The core of future development lies in promoting the three models from imbalance to synergy,and building a stratified,flexible,and institutionalized co-governance system.