S0129Critical

分配性傷害與機會不平等

Operation & Monitoring

Risk Description

AI systems produce systematically unequal outcomes in resource allocation, opportunity access, or service provision decisions. For example, loan approval AI systematically granting lower amounts to applicants from specific postal codes, job screening AI having implicit preferences for graduates of specific schools, or insurance pricing AI charging higher rates for specific socioeconomic groups. These decisions directly impact individuals economic opportunities and quality of life.

Framework Mappings

iso 23894R3
tw principle公平與不歧視
tw principle問責
tw risk type社會系統性影響(Societal Systemic)

Controls English translation pending

  • 公平性指標監控
  • 決策審計

Implementation Steps English translation pending

  1. 定義並監控多維度的公平性指標(如均等化率、機會均等性)
  2. 建立自動化的分配結果審計系統,偵測群體間的系統性差異
  3. 對高影響決策(貸款、就業、保險)實施強制公平性測試
  4. 建立受影響者的申訴和決策覆審機制