Glossary

Plain definitions of the terms used across this guide, with links to where each one is explained.

AI activity: What a student does with an AI tool, such as asking it for an outline. Not the same as learning.; See AI and learning →
AI Assessment Scale: A five-level framework (No AI to AI Exploration) for stating how much AI use a task allows.; See Policy builder →
AI literacy, competency, fluency: Three stages of capability: understanding AI and its risks, applying it to defined tasks, and integrating it into complex work with independent judgment.; See AI and learning →
Assessment validity: Whether an assessment still measures the intended learning. With AI involved, ask whether the work shows what the student can do.; See Policy builder →
Augmentation vs automation: Using AI to support your own thinking, such as a tutor explaining a concept, versus having AI do the work for you.; See AI and learning →
Base rate: How common something is in a group. With rare misuse, a larger share of detector flags are false accusations.; See AI detectors →
Cognitive offloading: Handing mental work to a tool. Helpful for routine effort; harmful when it removes the thinking a student needs to practice.; See AI and learning →
Default rule: The institution’s rule on AI use that applies when a syllabus or assignment says nothing.; See Guideline types →
Delphi study: A research method that builds expert consensus through repeated rounds of feedback.; See Evidence →
Disclosure: A student’s statement of which AI tools they used, for what, and how they changed and checked the output.; See For students →
Educational purpose: Why AI is part of a task, such as practice, feedback or evaluation. A second dimension alongside the permission level.; See Policy builder →
False positive: A detector flagging human-written work as AI-generated.; See AI detectors →
Humanizer: A tool that rewrites AI text to avoid detection.; See AI detectors →
Hybrid text: Work combining human and AI writing, such as a human draft edited by AI. Detectors handle it poorly.; See AI detectors →
Independent competence: What a student can do alone, without AI. Different from performance with AI.; See AI and learning →
Model card: Documentation of an AI model’s capabilities, limits and evaluation.; See Policy artifacts →
NIST AI RMF: The U.S. National Institute of Standards and Technology’s AI Risk Management Framework: govern, map, measure and manage AI risk.; See Frameworks →
Preprint / working paper: A research paper shared before peer review. Useful early evidence, read with more caution.; See Evidence →
Service card: User-facing documentation of an AI tool: what it is, who it is for and the rules for using it.; See Policy artifacts →
System card: Documentation of how an AI tool is deployed at an institution: integrations, data flows and controls.; See Policy artifacts →
Universal Design for Learning (UDL): A framework for offering several ways to engage with material and show learning, so courses work for more students.; See AI and learning →
Use-case registry: An institutional list of approved and proposed AI uses.; See Policy artifacts →
Visibility label: A tag (Public, Institution-wide, Restricted, Confidential) saying who may see a governance document.; See Policy artifacts →