Foundation 5

Measurement and Limits

The proposed system depends on measuring contribution without pretending that causal inference is perfect.

Provenance auditing

The MPS foundations define provenance auditing as a probabilistic reconstruction of causal inputs using available data. The proposed Advantage Index then represents an estimated proportion of an economic return attributed to non-meritocratic inputs.

Contribution decomposition

The technical manuscript defines a qualifying contribution through three conditions: identifiability, materiality, and persistence. It models value over sets of interacting contributions rather than assuming each input acts independently.

Approximate marginal contribution

Because an AI training contribution may be embedded in model weights and interactions, the proposal uses perturbation-based estimates of the performance change associated with masking, changing, or down-weighting a contribution. These are estimators and therefore carry approximation error.

Interaction and time

The framework also models synergy between contributions and persistence over time. Direct and synergistic value can then feed a proposed royalty-flow equation whose policy parameters must respect a budget-balance constraint.

Epistemic humility

The foundations identify the counterfactual as the hardest problem: what would have happened without a particular advantage or contribution? The proposal therefore calls for proxy measures and for scaling interventions to the confidence of the evidence.