Appendix E. Glossary

AI agent: A system that can pursue a goal through multiple steps and may use external tools.

Algorithm: A defined procedure for solving a problem or performing a computation.

Artificial intelligence (AI): A broad field involving systems that perform tasks associated with capabilities such as prediction, perception, language, and planning.

Bias: A systematic pattern in data, design, or use that can produce skewed or unfair results.

Deep learning: Machine learning using neural networks with multiple layers.

Evaluation: Testing a system against defined measures and representative cases.

Foundation model: A model trained on broad data that can be adapted to many tasks.

Generative AI: AI that generates content such as text, images, audio, video, or code.

Hallucination: An AI-generated claim or output that is unsupported or incorrect but may sound plausible.

Inference: Using a trained model to generate a prediction or output.

Large language model (LLM): A model trained to process and generate language.

Machine learning: Methods that learn patterns from data to make predictions or support decisions.

Multimodal AI: AI that can process or generate more than one modality, such as text, image, audio, or video.

Prompt: An instruction or input provided to an AI system.

Prompt injection: An attempt to manipulate an AI system through malicious or misleading instructions in its input or retrieved content.

Retrieval-augmented generation (RAG): A method that retrieves external information and supplies it to a generative model as context.

Training data: Examples used to fit a model's parameters.

Validation data: Data used to tune or select a model during development.

Human oversight: A meaningful process through which people can review, question, intervene in, or stop an AI-supported workflow.