(This is the fifth in a series on AI in Emergency Medicine. You can find the previous posts here.)
If you are in a medical field and working with artificial intelligence, here is a list of terms you should know.
| Term | Plain-English Definition |
| ???? Algorithm | A clear set of instructions used by a computer to complete a task. Think of it as a recipe that solves a specific problem. |
| ???? Ambient AI | A tool that listens in the background and generates clinical documentation automatically. It turns doctor-patient conversations into notes, often in real time. |
| ???? Artificial Intelligence (AI) | Technology that allows computers to act in ways that seem smart. These systems can sort information, answer questions, or make suggestions based on patterns. |
| ???? Audit Trail | A record of the data, processes, and decisions made by an AI system, allowing for review and accountability. |
| ⚖️ Bias (in AI) | Patterns in AI responses that reflect unfair assumptions. These patterns may come from the data the system learned from and can affect results in subtle or serious ways. |
| ???? Cognitive Bias | A pattern of thinking that can lead to errors in judgment. In AI, it refers to the system’s tendency to make decisions based on flawed assumptions or data. |
| ⚙️ Automation | Technology that carries out tasks with little or no human effort. These tools can handle routine steps in documentation, scheduling, or decision support. |
| ???? Explainability | The ability of an AI system to provide understandable reasons for its decisions or predictions. This helps users trust and effectively use the system. |
| ???? False Negative | When a test or AI system incorrectly indicates that a condition is absent. This can lead to missed diagnoses or delayed treatment. |
| ???? False Positive | When a test or AI system incorrectly indicates that a condition is present. This can lead to unnecessary treatments or anxiety. |
| ????️ Integration | The process of incorporating AI tools into existing clinical workflows and systems to enhance efficiency and decision-making. |
| ???? Inference | The moment when an AI system makes a prediction or gives an answer based on what it has learned. It is the step where it applies what it knows. |
| ???? Interoperability | The ability of different AI systems and software applications to communicate, exchange, and interpret shared data effectively. |
| ???? Large Language Model (LLM) | A computer system trained to understand and generate human language. It has read a large amount of text and can now respond in sentences that sound natural. |
| ???? Machine Learning (ML) | A way for computers to learn by studying examples. Instead of following exact instructions, the system picks up patterns and makes decisions on its own after training. |
| ???? Neural Network | A system built to process information in layers. It is inspired by how the brain works. These systems are especially good at identifying patterns in complex data. |
| ????️ Natural Language Processing (NLP) | A method that helps computers work with written or spoken language. It includes reading, summarizing, translating, or answering questions. |
| ???? Predictive Analytics | The use of data, statistical algorithms, and AI to identify the likelihood of future outcomes based on historical data. |
| ⌨️ Prompt | The message or question you type into an AI system to get a response. This could be as simple as “Summarize this note” or “What are the causes of chest pain?” |
| ???? Structured Data | Information that fits neatly into tables or fields. This includes lab values, vital signs, or medication lists. |
| ???????? Supervised Learning | A type of machine learning where the system is trained using labeled examples. For instance, the system learns to identify pneumonia after being shown many chest X-rays labeled with or without pneumonia. |
| ???? Token | A piece of text used by AI systems during processing. It may be a word or part of a word. The system breaks down your input into tokens so it can work with the language more easily. |
| ???? Tool Fatigue | The exhaustion or overwhelm clinicians may feel when required to use multiple digital tools, potentially reducing the effectiveness of AI implementations. |
| ????️ Training Data | The examples that are used to teach an AI system. If the system is meant to identify fractures, its training data would include lots of labeled images of fractures. |
| ???? Validation | The process of testing an AI system to ensure it performs accurately and reliably in real-world scenarios. |
| ???? Unstructured Data | Information that does not follow a fixed format. This includes progress notes, dictated reports, or conversation transcripts. |
Last Updated on June 18, 2025
Sam Ashoo, MD, FACEP, is board certified in emergency medicine and clinical informatics. He serves as EB Medicine’s editor-in-chief of interactive clinical pathways and FOAMEd blog, and host of EB Medicine’s EMplify podcast. Follow him below for more…
