Multiple Systems Converging on Collective Suboptimum
多系統陷入集體次優
Operation & Monitoring
Risk Description
When recommendation systems across platforms continuously push more provocative content to capture user attention; due to unmitigated control gaps, each system's individual decisions align with its own objectives, yet the collective result is an across-the-board decline in the quality of the information environment, triggering compliance exposure and operational reputational costs.
Framework Mappings
NIST AI RMFGOVERN 1.1
ISO/IEC 23894§6.6
NIST AI 600-1Information Integrity
MAESTRO
ISO/IEC 5338運作與監控
MIT AI Risk RepositoryDomain 6
Risk Treatment & Implementation Guidance
Build collective externalities into platform mechanism design to avoid individual optimization degrading the information environment; Set collective constraints on recommenders—content-quality and diversity floors; Monitor cross-platform collective-suboptimal outcomes with ecosystem-level metrics