AI Glossary

The following definitions are adapted from WMed AI guidance and supplemented by authoritative sources including NIST, FDA, WHO, and federal regulatory frameworks.

  • Adaptive Learning
    Educational systems that adjust content and pacing based on learner performance.
    Source: WMed AI Use (Educational Applications)

    Agentic (Agentic AI)
    AI systems designed to act autonomously toward defined goals by making decisions, taking actions, and adapting based on context, often coordinating multiple steps or tools without constant human direction.
    Source: OpenAI; NIST

    Algorithm
    A defined set of rules or instructions used by a computer to solve a problem.
    Source: WMed AI Use; NIST AI Glossary

    API (Application Programming Interface)
    A method that allows different software systems to communicate and exchange data.
    Source: NIST

    Artificial Intelligence (AI)
    Computer systems designed to perform tasks requiring human intelligence such as reasoning and learning.
    Source: WMed AI Use; National Institute of Standards and Technology (NIST AI Risk Management Framework)

    Automation
    The use of technology to perform tasks with minimal human intervention.
    Source: WMed AI Use (Operational Efficiency)

    Bias (AI Bias)
    Systematic errors in AI outputs caused by skewed or incomplete data.
    Source: National Institute of Standards and Technology; World Health Organization AI Ethics Guidance

  • Clinical Decision Support (CDS)
    Tools that assist clinicians with recommendations to improve care decisions.
    Source: U.S. Food and Drug Administration (AI/ML SaMD guidance)

    Cloud Computing
    Delivery of computing services over the internet instead of local infrastructure.
    Source: National Institute of Standards and Technology Cloud Definition

    Computer Vision
    AI that interprets and analyzes images such as radiology or pathology scans.
    Source: U.S. Food and Drug Administration; WHO AI in Health

    Compliance
    Adherence to legal, regulatory, and institutional requirements.
    Source: WMed AI Use; HIPAA/FERPA frameworks

    Context Window
    The amount of information an AI model can consider at one time.
    Source: OpenAI / LLM documentation

    Data Governance
    Policies and processes ensuring data is accurate, secure, and appropriately used.
    Source: WMed AI Use; DAMA International (DAMA-DMBOK)

    Data Pipeline
    The process of collecting, transforming, and preparing data for analysis.
    Source: WMed Analytics context; Microsoft Data Architecture

    Data Quality
    The accuracy, completeness, and reliability of data.
    Source: DAMA-DMBOK

    Deep Learning
    A machine learning technique using multi-layer neural networks for complex analysis.
    Source: U.S. Food and Drug Administration; WHO

    De-identified Data
    Data stripped of personal identifiers.
    Source: HIPAA Privacy Rule

    Digital Twin (Healthcare)
    A virtual model of a patient or system used for simulation and prediction.
    Source: World Health Organization Digital Health

  • Embedding
    A numerical representation of data used to capture meaning and relationships.
    Source: OpenAI / vector database documentation

    Explainability
    The ability to describe how an AI system produces results.
    Source: NIST AI RMF; WHO AI Ethics

    FERPA (Family Educational Rights and Privacy Act)
    U.S. law protecting student education records.
    Source: U.S. Department of Education

    Fine-Tuning
    Adapting a pre-trained model with organization-specific data.
    Source: OpenAI; ML literature

  • Generative AI (GenAI)
    AI that creates new content such as text, images, or code.
    Source: WMed AI Use; EDUCAUSE

    Generative Adversarial Network (GAN)
    A model where two neural networks compete to generate realistic data.
    Source: ML academic literature

    GPT (Generative Pre-training Transformer)
    A class of large language models that are pre-trained on large datasets and use transformer architectures to generate human-like text, answer questions, and perform reasoning tasks.
    Source: OpenAI; NLP

    Guardrails
    Policies and controls ensuring safe and appropriate AI use.
    Source: WMed AI Use (Governance)

    Hallucination (AI)
    When AI generates incorrect or fabricated information.
    Source: OpenAI; WHO cautionary guidance

    HIPAA (Health Insurance Portability and Accountability Act)
    U.S. law governing protection of patient health information.
    Source: U.S. Department of Health and Human Services

    Human-in-the-Loop (HITL)
    Human oversight in reviewing AI outputs.
    Source: NIST AI RMF; WHO

  • Inference
    Using a trained model to generate predictions from new data.
    Source: ML standard terminology

    Internet of Things (IoT)
    Connected devices that collect and exchange data.
    Source: NIST; WHO digital health

    Interpretability
    The degree to which a human can understand model decisions.
    Source: NIST AI RMF

    Knowledge Graph
    A structured representation of relationships between concepts.
    Source: Academic AI research

    Large Language Model (LLM)
    AI trained on large text datasets to generate human-like language.
    Source: OpenAI; academic NLP research

  • Machine Learning (ML)
    AI systems that learn patterns from data.
    Source: FDA; NIST

    Metadata
    Data describing other data.
    Source: DAMA-DMBOK

    Model
    A trained AI system used for predictions or outputs.
    Source: ML standard terminology

    Natural Language Processing (NLP)
    AI that processes and understands human language.
    Source: FDA; WHO

    Neural Network
    A brain-inspired computational model used in machine learning.
    Source: ML standard terminology

  • Overfitting
    When a model performs well on training data but poorly on new data.
    Source: ML standard terminology

    Predictive Analytics
    Using data to forecast future outcomes.
    Source: WMed Analytics context; healthcare analytics standards

    Prompt
    The input or instruction given to an AI system.
    Source: OpenAI; WMed AI Use

    Prompt Engineering
    Designing prompts to improve AI outputs.
    Source: OpenAI best practices

    Protected Health Information (PHI)
    Individually identifiable patient health data.
    Source: HIPAA Privacy Rule

  • RAG (Retrieval-Augmented Generation)
    An AI architecture that combines information retrieval with generative models, allowing the system to pull relevant external data (e.g., documents, databases) to improve accuracy and reduce hallucinations in responses.
    Source: OpenAI; NLP 

    Reinforcement Learning
    A learning method based on rewards and penalties.
    Source: ML standard terminology

    Risk-Based Approach
    Evaluating AI systems based on their level of potential impact or harm.
    Source: NIST AI RMF; WHO AI governance

  • Sensitive Data
    Data requiring protection (PHI, student, financial, research data).
    Source: WMed AI Use

    Shadow AI
    Use of AI tools without institutional approval.
    Source: WMed AI Use (Governance & Risk)

    Shadow Data
    Data used outside approved systems.
    Source: WMed Analytics / governance context

    Simulation-Based AI
    AI used in training simulations such as virtual patients.
    Source: Medical education literature; AAMC

    Single Version of Truth
    A consistent, validated dataset used across an organization.
    Source: WMed Analytics Roadmap context

    Structured Data
    Data organized in predefined formats (e.g., tables).
    Source: Data management standards

    Supervised Learning
    Training using labeled datasets.
    Source: ML standard terminology

    Token
    A unit of text processed by an AI model.
    Source: OpenAI documentation

    Train (Model Training)
    The process of teaching an AI model by exposing it to data so it can learn patterns, relationships, and behaviors used to make predictions or generate outputs.
    Source: NIST; ML standard terminology 

    Training Data
    Data used to train an AI model.
    Source: ML standard terminology

    Transformer
    A neural network architecture that uses attention mechanisms to process and understand relationships in sequential data (such as language), forming the foundation of modern large language models like GPT.
    Source: OpenAI; NLP 

  • Unstructured Data
    Data without a fixed format (e.g., notes, documents).
    Source: Data management standards

    Unsupervised Learning
    Identifying patterns in unlabeled data.
    Source: ML standard terminology

    Validation (Model Validation)
    Testing a model to ensure accuracy on new data.
    Source: FDA AI/ML guidance

    Vector Database
    A database optimized for storing and searching embeddings.
    Source: AI architecture documentation

    Workflow Integration
    Embedding AI tools into existing systems (EHR, LMS, ERP).
    Source: WMed AI Use (Operational Use)