Glossary

Search results

AI Content Describer(also: NVDA AI Content Describer)

AI Content Describer is an add-on for the NVDA screen reader that uses multimodal AI models to generate descriptions of on-screen visual content — images, controls, icons, charts, and arbitrary screen regions — on demand. It gives NVDA users a JAWS-Picture-Smart-equivalent…

AI for Accessibility(also: AI4A, Artificial Intelligence for Accessibility)

An umbrella framing used by technology companies and researchers for applications of artificial intelligence — including computer vision, natural language processing, speech recognition, and generative models — intended to benefit disabled users. Common examples include…

AI-Generated Content(also: AIGC)

An umbrella term for text, images, audio, video, and other media produced by generative AI systems — especially large language models and diffusion-based text-to-image or text-to-video models — in response to user prompts. AIGC is widely used in creative tooling (backdrop…

Access-Stabilising Support

A design framing, introduced by Bhuiyan et al. (2026), that positions AI in Deaf education not as an autonomous translator or replacement instructor but as a mediated tool whose role is to preserve visual access, reinforce teacher-validated signs, and sustain comprehension…

Accessibility-Oriented Prompting(also: Accessible Prompting, A11y Prompting)

A prompt engineering strategy for large language models (LLMs) in which explicit accessibility requirements are included in the prompt when requesting code generation or UI design. Rather than relying on the LLM to infer accessibility needs from generic instructions,…

Anthropomorphism(also: Humanization, Anthropomorphization)

The attribution of human characteristics, emotions, intentions, or behaviors to non-human entities such as technology, animals, or objects. In assistive technology and conversational AI design, anthropomorphism raises important questions about how human-like an interface should…

Assistive AI(also: AI for Accessibility, Accessible AI, Accessibility AI)

Artificial intelligence systems designed specifically to support disabled people in performing tasks, accessing information, or navigating their environments. Examples include object recognition tools for blind users, automatic captioning for deaf users, and predictive text for…

Assistive Drone(also: Assistive UAV, Assistive Quadcopter)

A small unmanned aerial vehicle configured to assist a person with a disability — most often a blind or low-vision user — with tasks such as locating objects, navigating unfamiliar environments, scanning distant signage, and previewing walking-path conditions. Compared to…

Bixby Vision

Bixby Vision is a visual-assistance feature in Samsung's Bixby assistant, built into Samsung Galaxy phones, that uses the device camera and AI to describe scenes, read text, identify objects and currency, translate signs, and answer questions about the live camera view. For…

Chain-of-Thought(also: CoT, Chain of Thought Reasoning, Step-by-Step Reasoning)

Chain-of-thought is a prompting and model-design technique in which a large language model produces its intermediate reasoning steps before giving a final answer. Modern reasoning models (e.g., OpenAI o-series, Claude thinking modes) expose chain-of-thought as visible internal…

ChatGPT(also: GPT, OpenAI ChatGPT)

ChatGPT is a conversational generative AI assistant developed by OpenAI, based on the GPT family of large language models. Users interact via a text chat interface and, in newer versions, through voice, image, and file upload. ChatGPT is widely used as an accessibility tool —…

Context Leakage(also: chat session context bleed)

A failure mode in conversational AI systems in which information from earlier, unrelated chat sessions or turns influences the current response, producing output that blends contexts — for example, mixing VoiceOver (macOS) instructions into a JAWS (Windows) troubleshooting…

Conversational User Interface(also: CUI, Conversational Interface, Dialogue Interface)

A user interface that enables interaction through natural language conversation, either via voice (spoken dialogue) or text (chat). Conversational user interfaces encompass voice assistants, chatbots, and dialogue systems that interpret user intent and respond in natural…

Diffusion Model(also: Diffusion-based Generator, Denoising Diffusion Model)

A diffusion model is a class of generative AI that learns to produce images or videos by iteratively denoising a random noise input, reversing a forward process that gradually adds noise to training data. In accessibility work, diffusion models are used to synthesize sign…

Disability-Centered Evaluation(also: Disability-Centric Evaluation, Disability-First Evaluation)

An approach to evaluating AI systems, tools, or research artefacts that places disabled people's lived experiences, information needs, and failure contexts at the centre of study design — including which data are collected, how ground truth is annotated, which models are tested,…

Find My Things(also: FMT)

A Microsoft-developed smartphone app that lets people who are blind or have low vision train a personalized object recognizer by recording short example videos of their own items, then use the trained model to locate those items later via audio and haptic guidance. Find My…

Image Description App(also: Visual assistance app, AI visual description app)

A smartphone or wearable application that captures an image of the user's surroundings and returns a spoken or textual description of its content, aimed primarily at blind and low-vision users. Early crowdsourced systems such as VizWiz (2010) relied on remote human workers;…

LLM Accessibility(also: AI-Generated Accessibility, Generative AI Accessibility)

The study and practice of how large language models (LLMs) and generative AI tools handle digital accessibility in their outputs, particularly when generating code, user interfaces, or content. Research in this area examines whether LLM-generated code meets accessibility…

LLM Self-Reflection(also: AI Self-Assessment, Model Self-Evaluation)

A technique in which a large language model is prompted to evaluate and critique its own output, identifying errors, gaps, or areas for improvement. In the context of accessibility, LLM self-reflection involves asking the model to assess whether the code or UI it generated meets…

Low-Resource Sign Language

A sign language for which standardised corpora, training data, technical infrastructure, and institutional support are limited compared to 'high-resource' sign languages like American Sign Language (ASL) or German Sign Language (DGS). Low-resource sign languages — such as Bangla…

Material Culture

The physical objects, artefacts, and environments that a community produces, uses, and inhabits, along with the meanings and practices embedded in them. In AI and accessibility research, material culture matters because computer vision systems trained on objects from one…

Microsoft Copilot(also: Copilot, Microsoft 365 Copilot, Copilot in Excel)

Microsoft Copilot is a family of generative AI assistants integrated into Microsoft 365 applications including Excel, Word, PowerPoint, Outlook, and Teams, as well as GitHub and Windows. In Excel and Google Sheets-style workflows, Copilot lets users describe spreadsheet…

Momentous Depiction

A conceptual framework proposed by Niu, Clements, and Kim (2026) for using generative AI to visualize critical moments that convey the insights and meanings of disability in storytelling videos. The framework identifies four core GenAI affordances that support or constrain…

Music GenAI(also: Generative Music AI, AI Music Generation)

Generative AI systems that produce musical output — melodies, full songs, instrumental accompaniment, or vocal tracks — from text prompts, seed audio, or structured parameters. Examples include Suno, Udio, MusicLM, and MusicGen. In accessibility and therapy contexts, music GenAI…

Perceptual Gap

A design failure identified by Choudhury (2026) in which an AI system's explanation is delivered through exactly the sensory channel that its user cannot access. For example, a Grad-CAM heat map overlaid on an image tells a blind user where the model looked but cannot be seen by…

Picture Smart AI(also: JAWS Picture Smart AI)

Picture Smart AI is a feature of the JAWS screen reader (Freedom Scientific/Vispero) that uses multimodal AI models to describe images, charts, and on-screen content on demand. It can describe a photo, explain a chart, read text embedded in an image, or answer follow-up…

Product Identification(also: Product Recognition)

The task of determining what a packaged or unpackaged product is from visual (or other sensory) input, at a level of detail useful to an end user: generic type (soup, cereal, shampoo), brand (Campbell's, Kellogg's, Dove), and variety or flavour (tomato vs. chicken noodle; 90%…

Prompt Contradiction

A type of large language model failure in which the system disregards explicit instructions or constraints given in the user prompt, producing output that contradicts what was asked. For example, an AI responding with visual instructions like "click the green button" after the…

Referential Drift

Referential drift is a failure mode in AI-generated sign language where spatial loci — the established positions in signing space used to refer to people, objects, or locations — shift or are not maintained consistently across a sentence. Because signed languages use spatial…

Regional Sign Variation(also: Sign Language Dialect, Regional Sign Dialect)

Regional sign variation refers to systematic differences in the form of signs across geographic regions within a single sign language, analogous to dialects in spoken languages. Variation arises from local Deaf school traditions, contact between communities, and historical…

Sign Spotting(also: Sign Detection, Continuous Sign Spotting)

Sign spotting is the task of automatically locating instances of specific signs within a continuous signing video, as opposed to classifying a pre-segmented isolated sign. It is a building block for search-by-sign in archive footage, automatic captioning of signed media, and…

Story Completer

A design role for generative AI in storytelling, proposed by Niu, Clements, and Kim (2026), in which AI systems complete and enrich stories authored by human creators rather than generating full storylines or automating creative decisions. The concept is framed in contrast to AI…

Therapeutic Drift

A phenomenon in AI-mediated support contexts in which users begin to treat an emotionally supportive AI system as a substitute for professional clinical care or meaningful human relationships, rather than as a complementary tool. Therapeutic drift poses risks in caregiving and…

Visual Assistant(also: Visual Assistant Skill)

Proposed by Gonzalez Penuela et al. (2026), the 'visual assistant' skill is a set of behaviours that an AI visual interpretation system must exhibit — beyond simply producing accurate captions — to meaningfully support blind and low-vision users in daily life. The nine proposed…

Visual Interpretation(also: AI Visual Interpretation, Visual Interpretation Application)

The task of translating visual content — a scene, object, document, interface, or image — into a form accessible to a blind or low-vision user, typically spoken or written description, answers to questions, or actionable guidance. Visual interpretation systems may be…

35 results.