Trustworthiness Tradeoffs Question
How should organizations navigate tradeoffs between trustworthiness characteristics when optimizing for one necessarily degrades another?
The NIST AI RMF acknowledges specific tradeoff patterns:
- Interpretability vs. privacy (explaining decisions may reveal protected information)
- Predictive accuracy vs. interpretability (more complex models are harder to explain)
- Privacy-enhancing techniques vs. accuracy under data sparsity (differential privacy can reduce model performance)
The framework notes that these tradeoffs “depend on the values at play in the relevant context and should be resolved in a manner that is both transparent and appropriately justifiable.”
But it offers no methodology for how to make these tradeoff decisions, who decides, by what criteria, with what accountability.
This is where the “voluntary framework” model may hit its limits. Tradeoff decisions are inherently value-laden. Without external requirements, organizations will predictably optimize for characteristics that serve their interests.