A STARTING POINT
Common terms: UX, research & AI
Unfamiliar language should not be a barrier to getting started.
Brief explanations, practical examples and links to the sources behind them. These are our plain-English paraphrases, not verbatim definitions or a substitute for reading a source in context. Examples and cautions are editorial guidance.
15 terms · Sources checked 19 October 2026. Foundational material retains its original date at source.
15 of 15 terms
Design
Accessibility
Designing and developing products so people with disabilities can perceive, understand and operate them, and participate without avoidable barriers.
In practice: A form has clear labels, keyboard access, understandable errors and support for screen readers.
Keep in mind: Automated checks alone cannot establish that an experience is accessible.
Source: W3C WAI: Introduction to Web Accessibility (opens an external website)AI
Agentic AI · AI agents
Here, AI systems that use tools and take steps towards a goal, potentially acting in other services. The level of autonomy and the meaning of “agent” vary between sources.
In practice: An assistant compares conferences and drafts a plan, but needs permission before booking or sending messages.
Keep in mind: A marketing label does not establish competence or safe autonomy. Examine actual tools, permissions and recovery options.
Source: Anthropic: Building Effective AI Agents—industry engineering guidance (opens an external website)AI
Generative AI · GenAI
AI that generates content such as text, images, audio or code using patterns learned from data.
In practice: A tool generates alternative draft interface messages for a designer to assess.
Keep in mind: A plausible output is not necessarily accurate, suitable or supported by evidence.
Source: NIST: Generative artificial intelligence (opens an external website)AI
Hallucination / confabulation · AI output
Generated content that is erroneous or false but may be presented confidently or plausibly.
In practice: An AI summary cites a journal article that cannot be found because the reference was invented.
Keep in mind: Check claims against original sources. A confident tone is not a reliability measure.
Source: NIST: Generative AI Profile, section 2.2—Confabulation (PDF) (opens an external website)AI
Human oversight
Arrangements that give people responsibility and practical means to understand, review and intervene in an AI system’s use.
In practice: A researcher can inspect supporting evidence, reject a recommendation and stop a delegated action.
Keep in mind: Having a person nominally involved does not show they have the information, time or authority to exercise meaningful oversight.
Source: NIST AI RMF: AI Risk Management and Human–AI Interaction (opens an external website)Design
Human–computer interaction · HCI
An interdisciplinary field concerned with how people interact with computing systems, and how those systems are designed, studied and evaluated.
In practice: An HCI study might examine how a voice assistant supports people with different communication needs.
Keep in mind: HCI reaches beyond websites and apps into social, physical and cultural contexts.
Source: ACM SIGCHI: About the field and community (opens an external website)AI
Large language model · LLM
A language model trained at large scale to learn patterns in language and generate or process text. Some systems combine language with images, audio or other inputs.
In practice: A chat tool drafts a summary that a researcher checks against the original paper.
Keep in mind: Fluent writing does not establish that a statement is true or that a reference exists.
Source: Google: Machine Learning Glossary—large language model (opens an external website)Design
Prototype
A representation of a proposed design, made to investigate questions before committing to a finished product. It can range from paper sketches to interactive code.
In practice: Clickable screens help a team investigate whether people understand a proposed application flow.
Keep in mind: A polished prototype does not prove the design works; choose its level of detail to suit the question.
Source: GOV.UK Service Manual: Making prototypes (opens an external website)Research
Qualitative research
Research that examines experiences, meanings and processes through material such as observations, interviews and documents.
In practice: Analyse how participants explain their hesitation while trying a prototype.
Keep in mind: Depth of understanding does not automatically establish how common a pattern is in a wider population.
Source: GOV.UK: Usability testing—qualitative studies (opens an external website)Research
Quantitative research
Research that uses numerical measurements to describe patterns or test questions, with attention to how data were collected and analysed.
In practice: Measure task completion and time taken across a defined group of participants.
Keep in mind: Numbers do not automatically make a study representative or explain why something happened.
Source: GOV.UK: Qualitative and quantitative usability testing (opens an external website)Design
Usability
How well particular people can achieve their goals with something, considering success, effort and satisfaction in the context where they use it.
In practice: Can someone find the right bus ticket, complete payment and understand what to do next?
Keep in mind: Always state whose goals and which context you evaluated. General usability testing can overlook disability-related barriers.
Source: W3C WAI: Accessibility, Usability, and Inclusion (opens an external website)Research
Usability testing
Evaluating a design by observing people attempting relevant tasks. Sessions may also collect explanations, errors and performance measurements.
In practice: Ask someone to locate an archived newsletter without telling them which menu to use.
Keep in mind: Evaluate the design, not the person. Leading instructions can conceal problems.
Source: GOV.UK Service Manual: Using moderated usability testing (opens an external website)Design
User experience · UX
A person’s experience of a product or service across their interactions with it, including expectations, feelings and outcomes. It extends beyond the appearance of its interface.
In practice: Booking an appointment involves finding information, choosing a time, receiving confirmation and attending—not only using a booking screen.
Keep in mind: A visually polished interface can still sit within a frustrating service.
Source: Norman and Nielsen: The Definition of User Experience (opens an external website)Design
User interface · UI
The means through which a person interacts with a system. Interfaces can include screens, controls, sound, speech and physical devices.
In practice: A ticket machine’s buttons, screen, payment reader and spoken instructions are parts of its interface.
Keep in mind: UI and UX are related, but describe different things.
Source: NIST: User interface (opens an external website)Research
User research · UX research
Systematic enquiry into people’s needs, behaviour and circumstances to inform decisions about products and services.
In practice: Observe how people currently arrange childcare before proposing a new booking service.
Keep in mind: Team opinions and AI-generated accounts of imagined users are not observations of actual participants.
Source: GOV.UK Service Manual: Learning about users and their needs (opens an external website)