Name
Governing the AI-Generated Item Pool: Quality, Bias, and Provenance Controls at Scale
Description
AI makes item authoring faster - and quietly reshapes the risk profile of your item pool. Semantic drift, content homogenization across generated batches, and subtle bias slip past review workflows designed for human-authored items. This session presents a governance and validation framework from a production psychometric environment: construct-alignment intake checks, validation protocols for AI-originated items, batch-diversity safeguards, and provenance tracking. Attendees leave with a governance checklist they can adapt - whether they build, buy, or blend.
Speakers
Primary Topic
Design, Development, and Psychometrics
Session Area
Certification & Licensure (C&L) Division, Technology-Based Assessment Committee, Equity Committee