Rebound Effect Offsetting Efficiency Gains
效率提升引發的回彈效應
Re-evaluation
Risk Description
When an organization adopts efficient inference technology to lower unit compute costs; because the system lacks aggregate capping and justification reviews, lower costs stimulate unconstrained expansion of AI use cases (Jevons paradox), driving up total energy consumption and violating decarbonization targets.
Framework Mappings
NIST AI 600-1Environmental Impacts
ISO/IEC 42005影響評估流程
ISO/IEC 5338重新評估
MIT AI Risk RepositoryDomain 6
Risk Treatment & Implementation Guidance
Set aggregate caps against rebound effects so lower unit cost does not raise total usage; Review the necessity of expanded application via use-justification assessment; Monitor total energy so efficiency gains are tied to aggregate targets