Cut-score panels using the Modified Angoff method ask panelists to estimate the probability that a Minimally Competent Candidate (MCC) would answer each test item correctly — a concept operationalized when individual human cognition was the unspoken unit of analysis. In 2026, with AI tools integral to instructional practice, panelists face an unsurfaced fork on every item: are they estimating MCC performance with AI assistance, without it, or with some implicit mixture? Performance Level Descriptors (PLDs) rarely specify the assumption, producing credentials whose construct of competence is undefined at its boundary with AI. Written from inside an active International Board of Standards for Training, Performance and Instruction (IBSTPI) Certified Professional Instructor cut-score panel, this paper proposes a three-state model (AI-prohibited, AI-permitted, AI-required) specified in PLDs and disclosed on the credential itself.
Marolyn Deidre Machen (Mon,) studied this question.