Technology terms
A plain definition for each term a student is likely to meet this year, with the source and a tag for what the word names. The tags follow the first question on the Kinds of technology page: a technology is a capability like machine learning or a distributed ledger; a product is a system built from technologies, like a case management system; a concept is an idea or a critique, like the digital divide, not a product you buy.
Seventy-eight terms in seven groups. Eighteen have a Flag: a published finding about how the word is applied in practice, with a link to the source. A flag is not a verdict on the technology, and an entry without one has no published measurement behind it in either direction.
AI and data
-
Large language model technology
A language model trained with deep learning on very large text datasets.
-
Small language model technology
A language model much smaller than an LLM, optimized for efficiency; there is no agreed size threshold.
-
Foundation model technology
A model trained on broad data that can be adapted to many downstream tasks.
-
Generative AI technology
AI models that learn the patterns of their training data and produce new text, images, audio, or video from them.
Flag A 2025 analysis in Philosophy and Technology reports no unique or unambiguous technical definition of the term. Ronge et al. 2025
-
Predictive model technology
A model built from historical data to forecast or classify a future outcome.
-
Machine learning technology
Building systems that learn patterns from data rather than being explicitly programmed.
-
Deep learning technology
Machine learning using many-layered neural networks.
-
Neural network technology
A computing system of interconnected nodes loosely modeled on biological neurons.
-
Transformer technology
A neural-network architecture that uses attention to process sequences; the basis of modern language models.
-
Fine-tuning technology
Adapting a pretrained model to a specific task with further training.
-
RAG technology
Retrieval-augmented generation: combining a search step with a generative model to ground its answers.
-
AI agent technology
Software that senses its environment and takes actions on its own to reach a goal it was given.
Flag A 2025 AIES paper argues the term has been diluted beyond utility and needs redefining. Bent 2025, AIES
-
Multimodal AI technology
A system that takes in and combines more than one data type, such as text and images.
-
Computer vision technology
Systems that interpret images or video.
-
Natural language processing technology
The field of getting computers to work with human language.
-
Open-weight model technology
A model whose trained parameters are published, so it can be inspected and fine-tuned; not the same as open source.
-
Training data concept
The dataset a model is fitted on; its flaws become the model’s.
-
Hallucination concept
A generative model producing fabricated content presented as fact.
-
Prompt engineering concept
Designing the input to a model to steer its output.
-
Federated learning technology
Training one shared model across several organizations by exchanging model updates, so the records never leave each site.
-
Synthetic data technology
Artificial records built from real data so the overall patterns hold, while no single row belongs to an actual person.
-
Red teaming concept
A planned test in which people deliberately try to make a system fail or behave badly, before it reaches the public.
Flag A 2024 AIES survey found red-teaming practice diverging on purpose, on what is evaluated, and on what the result decides. Feffer et al. 2024, AIES
Immersive and spatial
-
Virtual reality technology
A computer-rendered visual and audio scene built to imitate real-world sensation as the user moves within it.
-
Augmented reality technology
Digital content mixed into the real world in real time, positioned so it appears anchored to real places.
Flag A 2022 review found no consensus definition, with virtual reality and projection systems also called AR. Radu et al. 2022
-
Mixed reality technology
Augmented reality in which virtual objects are placed so that they appear to be part of the physical scene.
Flag Interviews with 10 experts and a review of 68 papers identified six competing notions of the term. Speicher et al. 2019, CHI
-
Extended reality concept
The umbrella term for combined real and virtual environments, covering virtual, augmented, and mixed reality.
Flag A 2026 systematic review found industry definitions diverging systematically from academic ones. Han et al. 2026
-
Digital twin product
A digital representation of a real-world thing or process; NIST states there is no agreed definition of the term.
Flag Of 149 studies using the term, 87.9 percent applied it to models that did not meet the NASEM definition. npj Digital Medicine 2025
-
Spatial computing concept
Interaction in which the machine tracks and acts on real objects and spaces around the user; no standards body defines it.
-
Haptics technology
Touch-based interaction, covering skin sensation and the sense of force in muscles, joints, and tendons.
Sensing and devices
-
Wearable device product
A small networked electronic device worn on the body that senses or displays information.
-
Smartwatch product
A wrist-worn device that extends a phone, shows notifications, and tracks activity; no standards body defines the category.
Flag The FDA warns that smartwatches and smart rings from dozens of companies do not measure blood glucose. FDA safety communication
-
Remote patient monitoring product
A care model where a connected home device sends a patient’s health data to a provider.
-
Passive sensing technology
Continuous collection of behavioral or physiological signals from a phone or wearable, with no active input.
-
Digital phenotyping concept
Quantifying a person’s behavior and mental state from their own device data, moment to moment.
-
Internet of Things concept
Everyday objects networked to collect and exchange data.
Flag A review of 164 articles published between 2005 and 2019 identified 100 distinct definitions. Goumagias et al. 2021
-
Edge computing concept
Processing data near the network edge where the devices are, instead of sending it to a distant data center.
-
RFID technology
Radio tags and readers that identify objects at short range without line of sight.
-
Geofencing technology
Drawing a virtual boundary around a place and triggering actions on entry or exit.
Infrastructure and platform
-
Cloud computing technology
On-demand access to shared computing resources over a network.
-
On-premise concept
Computing hardware an organization owns or leases and runs in its own data center, instead of a provider's cloud.
-
API technology
A defined interface that lets one program call functions in another.
-
Platform product
A service or infrastructure that mediates between groups of users; one of the vaguest words in tech.
Flag A 2024 systematic review collected 64 definitions of the term from 58 scholarly sources. Zeamari and Laurier 2024
-
Interoperability concept
The ability of different systems to exchange and use each other’s data.
-
FHIR technology
A modern standard for exchanging healthcare data over web technologies.
-
Blockchain technology
A tamper-evident digital ledger shared across a network with no central authority.
Flag Of 82 companies that shifted to a blockchain identity, most showed no structural change beyond the rename. Akyildirim et al. 2020
-
Quantum computing technology
Computing with qubits that exploit superposition and entanglement; still early-stage.
-
Zero trust concept
A security approach that checks every request on its own terms, assuming the network is already compromised.
Flag The same NIST publication lists the absence of a common framework and vocabulary as a deployment gap, and states no single vendor supplies zero trust. NIST SP 800-207 app. B
-
Multi-factor authentication technology
Signing in with two different kinds of proof, such as a password plus a code sent to a phone.
-
Ransomware technology
Malicious software that locks an organization out of its own files and demands payment, sometimes also threatening to publish them.
-
Phishing concept
A fraud in which an email or website impersonates a trusted organization to obtain sensitive information.
Health and human services
-
Telehealth concept
Delivering clinical care and health services at a distance through electronic means.
-
Digital therapeutic product
Software that delivers an evidence-based treatment, distinct from a general wellness app.
Flag Of 79 products reviewed in JAMA Internal Medicine, 34 percent had no published efficacy or effectiveness study. Amyx et al. 2024
-
Electronic health record product
A digital patient chart accessible to authorized clinicians.
-
Clinical decision support product
A tool that delivers patient-specific guidance to a clinician at the point of care.
-
Case management system product
Software supporting assessment, planning, and coordination of a client’s services; a product, not a single technology.
-
Chatbot product
Software that simulates conversation using scripted rules or a language model.
-
AI scribe product
Software that listens to an encounter and drafts the clinical note for review.
Access and accessibility
-
Assistive technology technology
Equipment and the services around it that help a person with a disability do everyday tasks independently.
-
Screen reader technology
Software that speaks aloud, or sends to a braille display, what is on the screen, so a blind user can operate a device.
-
WCAG concept
The W3C standard of testable criteria for making web content usable by people with disabilities, cited by many laws and contracts.
Flag The FTC ordered accessIBe to pay $1 million over claims that its automated tool could make websites WCAG-compliant. FTC 2025
-
Plain language concept
Writing a reader can understand the first time through, organized and worded for the audience who has to act on it.
Governance and critique
-
Algorithmic bias concept
Systematic unfairness in a system’s output that disadvantages a group.
-
Explainability concept
A system giving human-understandable reasons for its output.
-
Model card concept
A short document reporting a model’s intended use, performance across groups, and limits.
-
Algorithmic audit concept
A systematic evaluation of an AI system for bias, harm, or failure to meet stated values.
-
Automation bias concept
People over-trusting an automated aid, causing errors of omission and commission.
-
Technosolutionism concept
Treating "add technology" as the default answer to a social problem.
-
Digital divide concept
The gap in access to and skill with technology, and the disparities it produces.
-
Surveillance capitalism concept
A business model that extracts behavioral data to predict and shape future behavior for profit.
-
Data sovereignty concept
The principle that data about a people is governed by that people’s own rules.
-
Privacy by design concept
Building privacy protections into a system’s architecture and defaults from the start.
-
Interpretability concept
A model is interpretable when a person can follow how it turns inputs into a decision without a second tool to explain it.
Flag A widely cited 2018 paper reports the term applied under several non-overlapping notions, often asserted without argument. Lipton 2018, CACM
-
AI safety concept
Work aimed at keeping AI systems from endangering human life, health, property, or the environment under their intended conditions.
-
AI governance concept
The rules, routines, and review steps an organization uses so its AI matches its own values, its policies, and the law.
Flag The authors of that definition report few prior attempts to define the term and quote an overview calling the area unorganized. Mantymaki et al. 2022
-
Data minimization concept
Collecting and keeping only the personal information needed for the stated purpose, and nothing beyond it.
-
De-identification technology
Removing or altering the details in a record that would let someone tell which person it came from.
Flag The FTC alleged Avast claimed to strip identifying information yet sold browsing data that included per-browser identifiers, full histories, and city-level location. FTC 2024
-
Differential privacy technology
Adding measured statistical noise to published results so that no single person’s record changes what the numbers show.
Flag A survey of deployed systems reports no consensus on how to set the privacy parameter that decides how much protection is actually applied. Dwork et al. 2019
-
Vendor lock-in concept
When moving off a supplier costs so much in data conversion, retraining, and downtime that an organization stays by default.
No term matches that search.
Self-check
Nothing here is graded, and answers remain in this browser until you press reset.
Ten items, alternating between a definition to name and a term to define. Each takes one answer and cannot be changed once it is checked.
Definitions and sources are from a July 2026 review; each link was checked then. Several standards bodies publish as PDFs that refuse automated requests; those links open normally in a browser. For the AI family in more depth, see Kinds of AI systems.