Algorithm Audit is a non-profit group that helps people understand what makes algorithms fair.
Using its ‘algoprudence’ methodology, it looks at real, urgent ethical issues. It gathers independent experts to give advice when an algorithm might be treating people unfairly, and it tests tools that can find and reduce bias. To decide what “fair AI” means in each case, it combines human judgment about right and wrong with the ability of statistics to check large amounts of data.
Algorithm Audit will develop and test a practical evaluation framework for Large Language Models (LLMs) and chatbots that interface between governmental bodies and citizens. They will form an independent advisory commission to explore the normative questions that could arise when a chatbot is supplying information to the public.
Algorithm Audit will expand its work on algorithmic non-discrimination through workshops, training, and case studies, translating value-based principles into practice for both publicly and privately used AI systems.
Algorithm Audit continued reviewing high-impact algorithms used in the private and public sector in the Netherlands. They distilled these cases into public advice reports for policymakers. They also developed and maintain open source AI auditing tools so others can detect AI bias in algorithms.
Algorithm Audit used its 'algoprudence' methodology to gather independent experts to give advice when an algorithm might be treating people unfairly, and tests tools that can find and reduce bias. To decide what "fair AI" means in each case, it combines human judgment about right and wrong with the ability of statistics to check large amounts of data.
