The Institute for Ethical AI & Machine Learning
The Institute for Ethical AI & Machine Learning is a Europe-based research center dedicated to developing frameworks that promote the responsible development, deployment, and operation of machine learning systems. Our cross-functional team of volunteers includes leaders from technology, machine learning, industry, policy, and academia (STEM, Humanities, and Social Sciences).
Our Commitment to Responsible AI
We are a research center conducting highly technical, practical, and cross-functional research across the 8 Machine Learning Principles. We collaborate with industry, academia, and governments to create frameworks and libraries that align with our 4-phase approach to responsible AI.
The Institute's 4-Phase Strategy
Our strategy for responsible AI development comprises four phases:
- By Principle: Empowering individuals through best practices and applied principles.
- By Process: Empowering leaders through practical industry frameworks and applied guides.
- By Standards: Empowering entire industries through contributions to industry standards.
- By Regulation: Empowering entire nations through our work.
The 8 Machine Learning Principles
Our Machine Learning Principles, developed by domain experts, offer a practical framework guiding technologists in the responsible development of machine learning systems. Below is a summary; for full descriptions, visit the principles page.
- Human Augmentation: Assessing the impact of incorrect predictions and designing systems with human-in-the-loop review processes where reasonable.
- Bias Evaluation: Continuously developing processes to understand, document, and monitor bias in development and production.
- Explainability by Justification: Developing tools and processes to improve transparency and explainability of machine learning systems where reasonable.
- Reproducible Operations: Developing infrastructure for a reasonable level of reproducibility across ML system operations.
- Displacement Strategy: Identifying and documenting relevant information to mitigate the impact of automation on workers.
- Practical Accuracy: Developing processes to ensure accuracy and cost metric functions align with domain-specific applications.
- Trust by Privacy: Building and communicating processes that protect and handle data with stakeholders who interact with the system directly or indirectly.
- Security Risks: Developing and improving processes and infrastructure to ensure data and model security during the development of machine learning systems.
Extended descriptions with case studies and examples for all principles are available on the principles page.
The AI-RFX Procurement Framework
The AI-RFX procurement framework provides templates to improve AI safety, quality, and performance in industry. This open-source framework translates the Principles for Responsible Machine Learning into a checklist. More information and download instructions are available on the AI-RFX Procurement Framework page.
Raising the Bar for AI
Developed by domain experts, the AI-RFX framework promotes best practices in the procurement, design, development, and integration of machine learning systems. It goes beyond AI algorithms, assessing the maturity of processes and technical infrastructure. The framework includes a request for proposal template and an assessment criteria template based on our Machine Learning Maturity Model (downloadable from the AI-RFX Procurement Framework page).
About The Institute
We are a Europe-based think tank uniting technology leaders, policymakers, and academics to develop industry standards for Data Governance and Machine Learning.
Ethical ML Network (BETA)
The Ethical ML Network (BETA) is a global network of engineers, scientists, managers, leaders, and thinkers who support the 8 principles for responsible machine learning and the 4 phases toward responsible AI development. The network is currently in beta. Join if you are:
- An AI startup/scale-up founder
- An industry professional
- A professor or academic
- An engineer
- A data scientist
- A product, project, or delivery manager
The Ethical ML Network (BETA) emphasizes that ethical machine learning requires responsible individuals advocating for best practices. This is reinforced in each of the Machine Learning Principles.
