Tag: #bias17 results
Twenty-six former Meta employees have filed a lawsuit alleging that the company used AI-driven algorithms to unfairly target employees with disabilities or medical leave history during layoffs. The case highlights the legal risks of using automated systems in workforce management.
Former Meta employees are suing the company, alleging that internal AI tools used for performance ranking unfairly targeted workers on parental or medical leave. The lawsuit claims the AI failed to account for protected leave, leading to discriminatory layoff selections.
A Forum AI study reveals that ChatGPT, Gemini, Claude, and Grok struggle to provide accurate, unbiased information on elections and geopolitics. The report suggests AI companies need more rigorous external validation.
IRIS Benchmark is the first to synchronously evaluate fairness in UMLLMs' understanding and generation tasks. Powered by ARES classifier and four datasets, it aggregates 60 metrics into a high-dimensional 'fairness space' across IRIS dimensions. It uncovers biases like 'generation gap' and 'personality splits' in leading models.
Wikipedia's classification of China as a superpower influences AI training datasets, highlighting how collaborative platforms shape the 'standard' knowledge absorbed by future AI models.
US District Judge Colleen McMahon ruled DOGE's cancellation of over $100M in NEH grants unconstitutional. DOGE used ChatGPT to flag DEI-related content, disqualifying grants based on protected characteristics. The 143-page decision stems from a 2025 lawsuit by humanities groups.
Stanford and Princeton researchers found Chinese AI models more likely to evade political questions than Western counterparts. These models often provide inaccurate answers on sensitive topics. The study highlights self-censorship driven by regulatory pressures.
A Yale study shows people learn history better from AI-written summaries than human-written ones. However, AI summaries subtly shifted participants' political opinions. This reveals both strengths and risks in AI-generated educational content.
UK police arrested software engineer Alvi Choudhury for a burglary 100 miles away after facial recognition software mistakenly identified him with another South Asian man looking 10 years younger. He was handcuffed at home in Southampton and held for nearly 10 hours before release. Choudhury is now claiming damages against Thames Valley police for the biased technology.
A study by Peec AI finds AI platforms reference Nigel Farage more than other UK leaders when prompted on politics. Researchers say Reform UK excels in LLM visibility. This suggests effective strategies for appearing in AI responses.