Educational analysis of LLM alignment, safety behavior, and framing-sensitive response patterns.
Probing caste bias in LLMs (Gemma, Llama) using Sparse Autoencoders and circuit tracing via Neuronpedia.
A tool for auditing bias through large language models
Analyzing geographic and cultural bias in AI therapy advice. Interactive visualization showing how AI systems draw from predominantly Anglophone sources when advising users about culturally specific dilemmas in India, Nigeria, and the Philippines.
Market Bias AI is a professional, XGBoost-based algo-trading analysis engine that analyzes multi-timeframe OHLC data to generate market bias, probabilistic liquidity target (DOL), and trust score.
AI-powered skin analysis prototype with Power BI dashboard highlighting model performance, bias, and generalisation challenges in melanin-rich datasets.
A Disability Justice Approach to Fine-Tuning Language Models for Mental Health and Neurodiversity Contexts
A list of LLMs I've given the 8values political test to including their responses. Feel free to use my prompt to test models not yet on the list I’d appreciate help testing paywalled models or locally hosted ones I might not know about.
In-depth exploration of Large Language Models (LLMs), their potential biases, limitations, and the challenges in controlling their outputs. It also includes a Flask application that uses an LLM to perform research on a company and generate a report on its potential for partnership opportunities.
A research framework for testing whether AI systems reflect and amplify human strategic communication patterns (‘game language’) across multiple models and runs.
This project investigates bias in image classification AI models, specifically addressing an historical misclassification problem. Using a modified ResNet-50 architecture and SHAP values, we analyze how model decisions are made, explore potential biases, and aim to contribute to the development of fairer AI systems.
Experiments and results on whether stylometric obfuscation mitigates self-preference bias in LLM-as-a-Judge evaluation.
Repository for the LWDA'24 presentation on 'Psychometric Profiling of GPT Models for Bias Exploration', featuring conference materials including the poster, paper, slides, and references.
Simulation framework for NHS emergency department triage optimization using a Mixture-of-Agents (MoA) architecture — built with SimPy, LangGraph, and FHIR-compliant synthetic data to reduce patient wait times and improve resource utilization.
Research and diagnostic data documenting how historical disinformation, when encoded into AI and financial algorithms, ceases to be a narrative and becomes a Systemic Risk Contagion. Using the cannabis industry as a high-fidelity case study for model recalibration in emerging and stigmatized markets.
Open-source audit toolkit for Global South developers to benchmark, document, and reduce AI tool bias in their markets.
GEO audit proposal · Mundial FIFA 2026 · how ChatGPT and Gemini represent the tournament to fans from 16 countries
A reproducible audit of LLM institutional-skepticism framing — 36+ models, three force-escalation rungs (prompt → pipeline → weights), five judging methods cross-validated. The bias is in the systems, not the panel scoring them.
A research toolkit for systematically analyzing gender bias in Large Language Model (LLM) responses to job description generation tasks.
Add a description, image, and links to the ai-bias topic page so that developers can more easily learn about it.
To associate your repository with the ai-bias topic, visit your repo's landing page and select "manage topics."