---
id: "action-use-mixture-weights"
type: "action-item"
source_timestamps: ["§ Ask the Bot", "§ The Music Industry Precedent"]
tags: ["data-valuation", "model-training"]
related: ["concept-data-mixture-weights", "framework-cmo-compensation", "question-weight-verification"]
action: "Extract and apply data mixture weights from the training process to determine the relative value of data sources."
outcome: "A low-cost, accurate distribution metric for dividing compensation among different classes of content creators."
audience: ["policymakers", "economists"]
sources: ["tail1"]
sourceVaultSlug: "hbr-seg-tail1"
originDay: 1
articleStem: "hbr-tail-109-ai-pay-fair-rates-content"
sourceUrl: "https://hbr.org/2026/06/how-ai-companies-can-pay-fair-rates-for-the-content-they-need"
sourceTitle: "How AI Companies Can Pay Fair Rates for the Content They Need"
---
# Use data mixture weights to calculate relative data value

## Action

**Extract and apply [[concept-data-mixture-weights]] from the training process to determine the relative value of data sources.**

Instead of pricing individual pieces of data post-hoc, policymakers and economists should use the proprietary mixing weights reported by model builders during training as a credible, mathematically sound signal of the relative value of different data categories — the mechanism behind **Step 2** of the [[framework-cmo-compensation]].

## Expected outcome

A low-cost, accurate distribution metric for dividing compensation among different classes of content creators.

## Dependency / risk

Hinges on being able to trust or verify reported weights — see [[question-weight-verification]].