Resource Hub
Explore guides, research papers, policy briefs, and tools to aid decisions on standards and assessments.
The “Too Good” Problem
Artificial intelligence can now predict item difficulty before students respond, potentially reducing pretesting costs and accelerating item development. But when should states trust these predictions, and what risks require management? This panel brought together assessment researchers and vendors to discuss practical realities of AI-based difficulty prediction. A state assessment director moderated, focusing on questions state leaders face when evaluating AI solutions: validation requirements, fairness concerns, appropriate applications, and cost–benefit tradeoffs.
