Measure agreement and adjudicate disputed labels using: What should happen before training proceeds?
Frequent, category-specific disagreement requires measuring agreement and clarifying labels before training.
The question
A retail demand-planning dataset contains product labels from several annotators. They disagree frequently about promotional events, and disagreement is concentrated in a new product category. What should happen before training proceeds?
Preparing for AIGP? Take the free 5-min readiness quiz →
- Measure agreement and adjudicate disputed labels using clarified guidance. ✓Agreement analysis reveals label uncertainty, while guidance and adjudication create a more consistent basis for training in the new category.
- Select the most experienced annotator’s labels as the definitive training target.Experience may inform adjudication, but unilateral selection does not test disagreement or document a shared labeling standard.
- Remove the new category until its sales history becomes longer.More sales history may help later, but it does not clarify current annotation definitions or disagreement sources.
- Train on all labels and average annotator judgments for each product.Averaging categorical judgments can obscure meaningful disagreement and does not establish which interpretation fits promotional demand.
The trap
When annotators disagree, improve the labeling rule and document adjudication before treating labels as ground truth. How to remember it
Frequent, category-specific disagreement requires measuring agreement and clarifying labels before training.
How many of these would you get right?
One of 1581 AIGP questions on Certsqill. Take a free five-minute check and see your score per domain — not one number, but which section to open tonight.
Test your AIGP readiness — freeMore Understanding How to Govern AI Development questions
- Conduct security testing for prompt injection: Which evidence gap should receive priority? →
- Version the data and model artifacts: What documentation control is most immediately required? →
- Test representative languages and severe safety scenarios: What is the decisive next step before relying on →
- All 426 Understanding How to Govern AI Development questions →
Part of the Certsqill AIGP question bank · Understanding How to Govern AI Development ·
Every answer, right and wrong, comes with its own explanation.