| Term | Defintion | Tier |
|---|---|---|
| AI Agent / Agentic AI | An AI-enabled system that can interpret goals, plan steps, use tools and take actions with varying levels of human involvement. (i.e., general purpose to act on request). | Tier 1: Core Terms |
| AI Alignment | The effort to make an AI system’s objectives and behavior consistent with intended values, rules and goals. | Tier 3: Advanced Terms |
| AI Bot/Assistant | An AI-powered program designed to answer questions, provide information, or perform specific predefined tasks in response to user requests (i.e., general purpose to respond to request). | Tier 1: Core Terms |
| AI Governance | Policies, roles, processes and controls used to guide and oversee AI decisions and use across the university. | Tier 2: Additional Terms |
| AI Governance Committee (offering current and future guidance) | University body responsible for institutional oversight of AI adoption and designated AI use cases, including alignment with mission, ethics, legal obligations and risk posture. | Tier 2: Additional Terms |
| AI Guardrail | A rule, technical control or process intended to limit unsafe, unauthorized or inappropriate AI inputs, outputs or actions. | Tier 1: Core Terms |
| AI Literacy | Knowledge and skills needed to understand AI capabilities and limitations, use AI effectively, evaluate outputs critically and apply it responsibly. | Tier 1: Core Terms |
| AI Model | A computational system trained on data to identify patterns and produce predictions, classifications, recommendations or generated content. | Tier 2: Additional Terms |
| AI Output | Content, analysis, recommendations, predictions or actions produced by an AI system. | Tier 2: Additional Terms |
| AI Risk Assessment | A structured review of an AI tool or use case to identify potential impacts and determine appropriate safeguards. | Tier 2: Additional Terms |
| AI Safety | Practices used to reduce the likelihood and impact of harmful, unreliable or unintended AI behavior. | Tier 2: Additional Terms |
| AI Task Force (offering targeted responses) |
University group charged with a point-in-time analysis to assess how emerging AI tools are used in teaching, research, operations, and student success/wellness and what Marquette's needs are related to AI. | Tier 2: Additional Terms |
| AI Tool or Use Case Assessment | University process used to evaluate a proposed AI tool or use case for privacy, security, compliance, accessibility, contractual terms and institutional fit. | Tier 2: Additional Terms |
| AI Tool | An application, feature or service that uses AI to perform or assist with a task (i.e., general purpose to help a user perform a task). | Tier 1: Core Terms |
| AI Use Case | An AI-enabled system, tool, bot, agent or automated function that operates on behalf of the university, uses institutional data, represents Marquette or incorporates university intellectual property. | Tier 1: Core Terms |
| Application Programming Interface (API) | A defined way for software systems to exchange data or request functions. | Tier 3: Advanced Terms |
| Approved AI Tool | An AI tool that has undergone appropriate university review and is acceptable for defined users, purposes and data sensitivity levels. | Tier 2: Additional Terms |
| Artificial Intelligence (AI) | Computer-based systems designed to perform tasks commonly associated with human intelligence, such as analysis, content generation, prediction or decision support. | Tier 1: Core Terms |
| Attention Mechanism | A method that helps an AI model weigh which parts of an input are most relevant. | Tier 3: Advanced Terms |
| Bias | A systematic tendency in data, model behavior, design or use that can produce unfair, distorted or unrepresentative outcomes. | Tier 2: Additional Terms |
| Data Privacy | Appropriate collection, access, use, sharing, retention and protection of information about individuals. | Tier 2: Additional Terms |
| Data Security | Administrative, technical and physical safeguards used to protect data and systems from unauthorized access, loss, alteration or disclosure. | Tier 2: Additional Terms |
| Data Sensitivity Level 1 | Category - Public: Information intended for public access or broad sharing where unauthorized disclosure would have little to no adverse impact on the university or individuals [Minimal impact] | Tier 2: Additional Terms |
| Data Sensitivity Level 2 | Category - Less Sensitive: Information that requires internal protection because unauthorized disclosure could cause reputational, operational or limited business harm to the university. [Moderate impact] | Tier 2: Additional Terms |
| Data Sensitivity Level 3 | Category - Sensitive: Information that requires restricted access because unauthorized disclosure could result in significant institutional, legal, compliance, financial or privacy risks. [Significant impact] | Tier 2: Additional Terms |
| Data Sensitivity Level 4 | Category - Critical: Information requiring the highest level of protection because unauthorized disclosure could result in severe legal, regulatory, financial, operational or personal harm. [Severe impact] | Tier 2: Additional Terms |
| Data Steward / Data Owner | University-designated authority responsible for governing the classification, use, sharing, protection and approval conditions for a defined set of institutional data. | Tier 2: Additional Terms |
| Deep Learning | A type of machine learning that uses multi-layer neural networks. | Tier 3: Advanced Terms |
| Deepfake | Synthetic or manipulated media that convincingly portrays a person saying or doing something that did not occur. | Tier 3: Advanced Terms |
| Embedding | A numerical representation of content that captures relationships or similarity. | Tier 3: Advanced Terms |
| Enterprise Data Protection (EDP) | Protections in an institutionally managed AI environment intended to prevent university prompts and data from being used outside the organization or to train public models. | Tier 1: Core Terms |
| Fine-Tuning | Additional training that adapts an existing model for a specific task, domain or style. | Tier 3: Advanced Terms |
| Foundation Model | A broadly trained AI model that can be adapted or used for many tasks. | Tier 3: Advanced Terms |
| Generative AI (GenAI) | AI that creates new content, including text, images, audio, video or code, in response to instructions or other inputs. | Tier 1: Core Terms |
| Grounding | Connecting an AI response to designated source information or data. | Tier 3: Advanced Terms |
| Hallucination | An AI-generated response that appears credible but is inaccurate, misleading or unsupported by reliable information. | Tier 1: Core Terms |
| Human Oversight / Human-in-the-Loop (HITL) | A control requiring a person to review, approve, correct or intervene in AI-supported work and remain accountable for final decisions. | Tier 2: Additional Terms |
| Information Sensitivity / Data Sensitivity Level | The assignment of information to Levels 1 through 4 based on potential impact, which determines the safeguards needed to protect it from inappropriate access, use or disclosure. | Tier 1: Core Terms |
| Institutional Data | Data classified under Marquette University’s information sensitivity framework and subject to university requirements for appropriate use, protection, sharing and governance. | Tier 2: Additional Terms |
| Knowledge Base | An organized collection of information that can be searched or used by people, applications or AI systems. | Tier 3: Advanced Terms |
| Large Language Model (LLM) | An AI model trained on large amounts of text to interpret and generate language-based content. | Tier 2: Additional Terms |
| Machine Learning (ML) | A branch of AI in which systems learn patterns from data rather than relying only on explicitly programmed rules. | Tier 2: Additional Terms |
| Model Context Protocol (MCP) | A technical protocol intended to standardize how AI applications connect to external tools, data and services. | Tier 3: Advanced Terms |
| Model Drift | A decline or change in model performance as data or conditions change. | Tier 3: Advanced Terms |
| Multi-Agent System | A system in which multiple AI agents coordinate or divide work. | Tier 3: Advanced Terms |
| Multimodal AI | AI that can work with more than one type of information, such as text, images, audio or video. | Tier 3: Advanced Terms |
| Natural Language Processing (NLP) | AI methods that enable computers to interpret, analyze and generate human language. | Tier 3: Advanced Terms |
| Neural Network | A machine-learning structure made of connected processing layers that learns patterns from data. | Tier 3: Advanced Terms |
| Prompt Engineering | The practice of structuring and refining prompts to improve AI responses. | Tier 3: Advanced Terms |
| Prompt | The instruction, question, context or information a person provides to an AI system. | Tier 2: Additional Terms |
| Public AI Tool | An AI service whose public or consumer version may not provide institutionally approved protections for university data. | Tier 2: Additional Terms |
| Responsible AI | The design, selection and use of AI in ways that are ethical, lawful, secure, transparent, accountable and aligned with institutional values. | Tier 1: Core Terms |
| Retrieval-Augmented Generation (RAG) | A method that retrieves relevant source material and supplies it to a generative AI model when creating a response. | Tier 1: Core Terms |
| Shadow AI | AI tools or systems used without appropriate university review, approval or oversight. | Tier 1: Core Terms |
| Synthetic Data | Artificially generated data designed to resemble characteristics of real data. | Tier 3: Advanced Terms |
| Synthetic Media | AI-generated or substantially AI-altered images, audio, video or other media. | Tier 3: Advanced Terms |
| Temperature | A model setting that influences how predictable or varied generated responses may be. | Tier 3: Advanced Terms |
| Token | A unit of text processed by a language model. | Tier 1: Core Terms |
| Training Data | Information used to teach an AI model to recognize patterns or produce outputs. | Tier 3: Advanced Terms |
| Transformer Model | A neural-network architecture widely used in modern language and multimodal models. | Tier 3: Advanced Terms |
