Glossary

Computer systems that perform tasks mimicking human intelligence, using algorithms and data. This includes understanding language, recognizing patterns, problem-solving, and decision-making, and may involve technologies like machine learning, neural networks, and robotics.

The tendency to over-rely on AI outputs even when they may be incorrect

Information technology applications that augment clinical reasoning and medical decision-making in the exam room

The phenomenon where a machine learning system underperforms due to a mismatch between the data on which the model was trained and the data on which it is deployed.

The process by which an overreliance on artificial intelligence systems can result in loss or degradation of essential human capabilities.

When AI, especially large language models (LLMs), generates incorrect or fabricated information that appears plausible. Essentially, the AI "makes things up" or presents false information as fact.

The uneven performance of AI, where it is highly capable in some tasks and poor in others

Advanced AI systems trained on massive texts to understand and generate language

A type of artificial intelligence that enables computers to learn from data without being explicitly programmed

Crafting inputs for useful and accurate results from AI tools

Any data that can identify a patient, such as name, date of birth, medical record number, or voice. This is defined by HIPPA, federal legislation. Based on this legislation, third-party entity that handle protected health information on behalf of a covered entity (healthcare provider, insurer) must have a formal agreement called a business associates agreement on how that data will be handled.

The information used to teach machine learning models how to recognize patterns, make decisions, and learn from data. It's a collection of labeled examples, like images with labels for objects or text with tagged sentences, that helps the AI algorithm understand the relationships between input and output.