Systemic vs. Individual AI Harms

AI governance frameworks typically focus on individual-level harms: unfair treatment, privacy violations, discriminatory outputs. These matter, but they miss systemic democratic threats.

Systemic harms operate through structural modalities, cascade failures, path dependencies, procedural erosion. They affect democratic institutions themselves, not just individuals within them. Examples: concentration of power among few AI providers creating infrastructure dependencies; emergent behaviors from interacting AI systems threatening institutional stability; systematic exclusion of entire communities from democratic participation.

Individual fairness metrics and group fairness constraints don’t capture these structural threats. A system can pass every fairness audit while still contributing to democratic fragility through the relationships and dependencies it creates.

Related: 05-molecule—democratic-risk-taxonomy