
AI FAIRNESS
GLOBAL LIBRARY
Resources

AI Fairness 360
This extensible open source toolkit can help you examine, report, and mitigate discrimination and bias in machine learning models throughout the AI application lifecycle. We invite you to use and improve it.
Produced mainly for: Tech teams
Origin: North America
Language: English
Type: Tech tool
Creator: IBM
%2019_10_15.png)
Audit-AI
Open sourced bias testing tool useful for developers, that seems active for updating
Produced mainly for: Tech teams
Origin: North America
Language: English
Type: Tech tool
Creator: Pymetrics

Bias and Fairness Audit Toolkit
Open source tech tool for auditing the data and predictions of machine learning solutions for their fairness. Requires relatively light technology background (can be used by non-developers)
Origin: North America
Language: English
Type: Tech tool
Creator: Chicago University

FairTest
Enables developers or auditing entities to discover and test for unwarranted associations between an algorithm's outputs and certain user subpopulations identified by protected features.
Produced mainly for: Tech teams
Origin: North America
Language: English
Type: Tech tool
Creator: Universidad de Columbia, Stanford, EPFL, Saarland Univ., Cornell Tech, Jacobs Institute

Themis™ o Themis-ML
A library that implements fairness-aware machine learning algorithms
Produced mainly for: Tech teams
Origin: North America
Language: English
Type: Tech tool
Creator: Massachusetts University

Responsible AI
The toolkit presents key risks associated to AI, including those related to bias & fairness. It offers a free responsible AI Diagnostic tool and a PDF version of pwc's Practical Guide to Responsible AI.
Language: English
Type: Guide
Creator: PwC
.jpg)
InterpretML
Open-source package for training interpretable (explainable) ML models
Language: English
Type: Tech tool
Creator: Microsoft

Responsible Innovation: A Best Practices Toolkit
Responsible Innovation is a toolkit for developers that provides a set of practices in development, for anticipating and addressing the potential negative impacts of technology on people. It covers applying Microsoft's AI principles of fairness, and approaches for harms modeling and community jury-style reviews.
Language: English
Type: Tech tool
Creator: Microsoft
.jpg)
Empowering AI Leadership – An Oversight Toolkit for Boards of Directors
A guide and toolkit for broader public and business leaders to consider not only AI fairness but also business strategy, governance and responsibility.
Language: English & Spanish
Type: Guide
Creator: World Economic Forum

TOOLBOX: Dynamics of AI Principles
The resource provides a list and a geographical worl map pointing ethical principle declarations and documents
Language: English
Type: Guide or manual
Creator: AI Ethics Lab

Algorithmic Accountability Policy Toolkit
This toolkit includes resources for advocates interested in or currently engaged in work to uncover where algorithms are being used and to create transparency and accountability mechanisms
Produced mainly for: Non - tech teams
Language: English
Tool: Guide or manual
Creator: AI Now Institute

RCModel, a Risk Chain Model for Risk Reduction in AI Services
The risk chain model (RCModel) supports AI service providers in proper risk assessment and control, and offers rpolicy recommendations
Origin: Asia
Language: English
Type: Guide or manual
Creator: The University of Tokyo

From Principles to Practice – An interdisciplinary framework to operationalize AI ethics
The paper offers concrete guidance to decision-makers in organizations developing and using AI on how to incorporate values into algorithmic decision-making, and how to measure the fulfillment of values using criteria, observables and indicators combined with a context dependent risk assessment.
Language: English
Tool: Guide
Creator: AIEI Group

Guidelines for Quality Assurance of Machine Learning-based Artificial Intelligence
The Guidelines for the Quality Assurance of AI Systems offers a comprehensive technical assessment of quality measures for AI systems, but it is not strictly speaking a document on AI Fairness. It is updated periodically in its original Japanese version, but an informal English translation is available too.
Language: Japanese
Type: Guide or manual
Creator: QA4AI
.jpg)
Fairness feature testing
.jpg)
CERTIFAI
Allows you to flag protected features in your dataset and then actively guides you through the selection of the best fairness metric to fit the specifics of your use case.
Produced mainly for: Tech teams
Language: English
Type: Tech tool
Creator: Data Robot
Tool developed by Cognitive Scale for data scientists to evaluate their AI models for robustness, fairness, and explainability, and allows users to compare different models or model versions for these qualities.
Language: English
Type: Tech tool
Creator: Cognitive Scale - Cortex

ML Fairness Gym
Open source development tool for building simple simulations that explore the potential long-run impacts of deploying machine learning-based decision systems.
Language: English
Type: Tech tool
Creator: Google
.jpg)
Ethics Canvas
Helps you structure ideas about the ethical implications of the projects you are working on, to visualize them and to resolve them.

Fairness-indicators: Tensorflow's Fairness Evaluation and Visualization Toolkit
Designed to support teams in evaluating, improving, and comparing models for fairness concerns in partnership with the broader Tensorflow toolkit. Perspective AI is provided as a content moderation case study.
Language: English
Type: Tech tool
Creator: Google
.jpg)
Ethically Aligned Design
Identifies specific verticals and areas of interest and helps provide highly granular and pragmatic papers and insights as a natural evolution of our work.
Produced mainly for: Tech teams
Language: English
Type: Guide or manual
Creator: IEEE Global A/IS Ethics Initiative
Produced mainly for: Managers
Language: English
Type: Tech tools
Creator: ADAPT Centre

Review into bias in algorithmic decision-making
It's more an educational publication than a tool (as the name suggests: "Review of.."). However, also provides some (high-level) recommendations for governments and regulators.
Origin: Europe
Language: English
Type: Guide
Creator: Centre for Data Ethics and Innovation

Data Ethics Charter
The IIF Data Ethics Charter outlines a set of principles for the ethical handling of customer data in the financial services industry and larger economy. Principles and examples of practice provide an overview of how financial institutions responsibly manage, protect, share, and use customer data.
Origin: Worldwide
Language: English
Type: Guide
Creator: Institute of International Finance (IIF)
.jpg)
AI Blindspot
A discovery process tool for product teams to spot unconscious biases and structural inequalities in AI systems.
Produced mainly for: Tech teams
Origin: Worldwide
Language: English
Type: Tech tool
Creator: Berkman Klein Center and MIT Media Lab
.jpg)
Responsible Data for Children
The work is intended to address practical considerations across the data lifecycle, including routine data collection and one-off data collections. It provides guidance, principles and practical case studies.
Origin: Worldwide
Language: English
Type: Guide, principles, case studies
Creator: UNICEF and GovLab

Model AI Governance Framework
Provides detailed and readily-implementable guidance to private sector organisations to address key ethical and governance issues when deploying AI solutions.
Produced mainly for: Managers & C levels
Origin: Asia
Language: English
Type: Guide
Creator: PDPC

AI For Everyone Course
Povide AI practitioners and non-technical professionals with the necessary tools to go all the way from foundational basics to advanced application, empowering them to build an AI-powered future.
Produced mainly for: For everyone
Origin: North America
Language: English
Type: Course
Creator: Andrew Ng

Explaining decisions made with AI
Mainly focused on explaining decisions made with AI, but it contains fairness issues in the model
Origin: Europe
Language: English
Type: Framework
Creator: Information Commissioner's Officer (ICO)
Guide of Algorithmic Auditing

An algorithmic audit identifies, addresses and corrects algorithmic bias so that you can act on informed and vetted AI-powered business recommendations.
Produced mainly for: Companies using AI
Origin: Europe
Language: English & Spanish
Type: Tech tool
Creator: Eticas

Principles for Accountable Algorithms and a Social Impact Statement for Algorithms
Aims to help developers and product managers design and implement algorithmic systems in publicly accountable ways. Accountability in this context includes an obligation to report, explain, or justify algorithmic decision-making as well as mitigate any negative social impacts or potential harms.
Produced mainly for: Managers & Tech teams
Origin: Worldwide
Language: English
Type: Guide or manual
Creator: Fairness, Accountability, and Transparency in Machine Learning Conference (FAT/ML)

The Case for Better Governance of Children’s Data: A Manifesto
Sets aspirational benchmarks to guide governments, the private sector and international organizations in developing data governance that take full account of children’s issues and rights, including in AI systems.
Origin: Worldwide
Language: English
Type: Guide
Creator: UNICEF

From Principles to Practice: use cases for implementing responsible AI in financial services
Findings from work by a group of partners in a project to explore the implementation of responsible AI principles, including Deutsche Bank, Linklaters, Microsoft, Standard Chartered, and Visa.
Produced mainly for: Managers & C levels
Origin: Asia
Language: English
Type: Guide
Creator: Microsoft, based on MAS FEAT principles
.jpg)
AI Principles
Promote use of AI that is innovative and trustworthy and that respects human rights and democratic values. Adopted in May 2019, they set standards for AI that are practical and flexible enough to stand the test of time.
Produced mainly for: Government and Ministerial Levels
Origin: Europe
Language: English
Type: Guide
Creator: OECD

Artificial intelligence - Governance framework model. Second Edition
Addition of industry examples in each section to illustrate how organizations have implemented AI governance practices.
Produced mainly for: Government
Origin: Asia
Language: English
Type: Guide
Creator: S Iswaran- Minister for Communications and Information- Singapore

¿Cómo implementar la debida diligencia en derechos humanos en el desarrollo de tecnología? El impacto en la privacidad
Helps identifying how the technological development may harm people's right to privacy. Either directly by the use of such technology, or indirectly as a result of the use of a third party.
Produced mainly for: Managers
Origin: Latin America and the Caribbean
Language: Spanish
Type: Guide
Creator: Asociación por los Derechos Civiles. Written by Leandro Ucciferri, with the independent consultants Agustina Bendersky and Denisse Cufré
.jpg)
Ethics & Algorithms Toolkit
A practical toolkit for cities to use to help them identify the risks of using an AI algorithm, and maps out the mitigating measures for different risks.
Origin: North America
Language: English
Type: Tech tool
Creator: GovEx, the City and County of San Francisco, Harvard DataSmart, Data Community DC
Model Artificial Intelligence Governance Framework Second edition

Incorporates the experiences of organizations that have adopted AI, and feedback from our participation in leading international platforms, such as the European Commission’s High-Level Expert Group and the OECD Expert Group on AI.
Produced mainly for: Manager
Origin: Asia
Language: English
Type: Guide or manual
Creator: SG:D, IMDA, PDPC (Singapore)
.jpg)
Fairness Toolkit (LiFT)
Linkedin - open source software library designed to enable the measurement of fairness in AI and machine learning workflows.
Origin: North America
Type: Tech tool
Creator: LinkedIn

Bias in machine learning and ethical implications
The report outlines ways in which financial institutions are cautiously implementing Machine Learning into their processes, such as in credit risk assessment and anti-money laundering, and the mathematical as well as social components of fairness to be considered.
Produced mainly for: C leve & Managers
Origin: North America
Language: English
Type: Guide
Creator: Institute of International Finance

Veritas Initiative
Framework for financial institutions to promote the responsible adoption of Artificial Intelligence and Data Analytics. It has developed a fairness assessment methodology in credit risk scoring and customer marketing, and has published whitepapers on the fairness assessment methodology and the open source code of these two use cases.
Produced mainly for: Tech teams
Origin: Asia
Language: English
Type: Tech tool
Creator: Monetary Authority of Singapore
Fair Pricing in Financial Services: summary of responses and next steps

Summarices the main themes in the submissions we received and, where appropriate, provide our responses. Provides further clarification on how we will apply our Framework in practice.
Origin: Europe
Language: English
Type: Guide
Creator: Financial Conduct Authority (FCA)

Responsible AI from Pilot to Production
A biased model that works for some users, and not others, is a failed model. Or a model that wasn’t sourced responsibly can be a poor reflection of company values and a nightmare with the media. It’s helpful to remember that AI reflects the people and the company that build it: when something goes wrong, it shows something may also be wrong internally.
Produced mainly for: C Level, Manager, Tech Team & Non - Tech Team
Origin: North America
Language: English
Type: Guide
Creator: Appen

The Impact of Ethics and Bias on Artificial Intelligence
AI is the study and research of providing software and hardware that attempts to emulate a human being. Cognitive computing is computing focused on reasoning and understanding that is inspired by human cognition. It is a subset of AI.
Produced mainly for: C levels & Managers
Origin: North America
Language: English
Type: Framework
Creator: David Schubmehl, Research Director, Cognitive/AI Systems
.jpg)
Data Privacy and Pou
The pou define why we exist, who we serve, what our goals are and how we make good decisions as kaitiaki for whānau.
Produced mainly for: Non - Tech teams
Origin: Pacific
Language: English
Type: Tech tool
Creator: Hiria Te Rangi and Amber Craig

Uso responsable de IA para política pública: manual de formulación de proyectos
This manual is part of a series of documents and tools developed by the fAIr LAC initiative to guide policy makers and their technical teams in mitigating the challenges inherent in AI-based decision support systems and promoting their responsible use.
Origin: Latin America and the Caribbean
Language: Spanish
Type: Manual
Creator: InterAmerican Development Bank (IDB)

Caja de Herramientas Humanísticas
Philosophical insights aimed at clarifying the nature of artificial intelligence, its relationship with human intelligence, as well as the various types of developments that have taken place in this field and the questions and challenges they raise.
Origin: Latin America and the Caribbean
Language: Spanish
Type: Guide
Creator: Guia.ai collaborating with CeTyS, BID & fAIr LAC

Privacidad y datos personales: una mirada desde el periodismo
To offer journalists and communicators basic notions about privacy and protection of personal data, the political and economic implications for democracy and the importance of preserving individuals' rights when reporting events of public interest.
Produced mainly for: Non - tech teams
Origin: Latin America and the Caribbean
Language: Spanish
Type: Guide
Creator: Asociación por los Derechos Civiles. Argentina-based NGO. Written by Xavier Ibarreche

Fairness Flow
Engages in cutting-edge research that can improve and power new product experiences at a huge scale for our community. Building on Facebook AI's key principles of openness, collaboration, excellence, and scale, we make big, bold research investments focused on building social value and bringing the world closer together.
Produced mainly for: Non - tech teams
Origin: North America
Language: English
Type: Guide or manual
Creator: Facebook

Ethical Toolkit for Engineering/Design Practice
Santa Clara University - Multi set of tools implementing ethical reflection, deliberation, and judgment into engineering and design workflows
Origin: North America
Language: English
Type: Tech tool
Creator: Markkula Center for Applied Ethics – Santa Clara University
%2011_10_37.png)
Model Cards
Short documents accompanying trained machine learning models that provide benchmarked evaluation in a variety of conditions, such as across different cultural, demographic, or phenotypic groups and intersectional groups that are relevant to the intended application domains.
Produced mainly for: C levels & Managers
Origin: North America
Language: English
Type: Guide or manual
Creator: Google
.jpg)
NZ Algorithm Charter
Demonstrates a commitment to ensuring New Zealanders to have confidence in how government agencies use algorithms. The charter is one of many ways that the government demonstrates transparency and accountability in the use of data.
Produced mainly for: Government agencies
Origin: Pacific
Language: English & Maori
Type: Framework
Creator: Aotearoa/NZ Government

Fairness Compass
Common definitions of fairness and ways of calculating the performance of a machine learning model. Mathematical tension across different fairness definitions makes it impossible to achieve "complete fairness." It helps stakeholders identify the most appropriate fairness definition for a specific use case via a decision tree.
Produced mainly for: Managers, Tech teams & Non - tech teams
Origin: Europe
Language: English
Type: Guide
Creator: Boris Ruf and Marcin Detyniecki (AI Research at AXA)
.jpg)
The Word Embeddings Fairness Evaluation Framework
Open source library for measuring bias in word embedding models. It generalizes many existing fairness metrics into a unified framework and provides a standard interface for: encapsulating existing fairness metrics from previous work and designing new ones, encapsulating the test words used by fairness metrics into standard objects called queries and computing a fairness metric on a given pre-trained word embedding model using user-given queries.
Produced mainly for: Tech teams
Language: English
Type: Tech tool
Creator: Millennium Institute for Foundational Research on Data (IMFD).

Real World AI: A Practical Guide to Responsible Machine Learning
This practical guide to deploying AI lays out a human-first, responsible approach that has seen more than three times the success rate when compared to the industry average.
Produced mainly for: C Level, Manager, Tech Team & Non - Tech Team
Origin: North America
Language: English
Type: Guide
Creator: Wilson Pang, CTO at Appen, and Alyssa Simpson Rochwerger, Director of Product at Blue Shield of California and Former VP of Data and AI at Appen
.jpg)
AI Ethics for Nonprofits toolkit
Participants will learn some of the fundamentals of AI ethics and then immediately get to practice applying ethical considerations related to the principle of Fairness in the context of several humanitarian and international development use cases
Produced mainly for: C Level, Manager, Tech Team, Non Tech Team
Origin: North America
Language: English
Type: Tool tech & Guide
Creator: NetHope (consortium of 60 INGOs) in partnership with MIT D-Lab and USAID

New Zealand research on Trust and Identity 2020
Is a purpose driven, inclusive, membership funded organization, whose members have a shared passion for the opportunities that digital identity can offer. Digital Identity NZ supports a sustainable, inclusive and trustworthy digital future for all New Zealanders. It is part of the NZ Tech Alliance.
Produced mainly for: C Level & Non Tech Team
Origin: Pacific
Language: English
Type: Guide
Creator: Digital Identity NZ
IA Responsable: Manual técnico: Ciclo de vida de la inteligencia artificial

The purpose of this manual is to provide technical recommendations and technical best practices in order to avoid contrary results (often unexpected) to the decision-makers' expectations.
Produced mainly for: Tech teams
Origin: Latin America and the Caribbean
Language: Spanish
Type: Manual
Creator: Felipe Gonzalez, Teresa Ortiz & Roberto Sánchez Ávalos for BID

Sesgo e Inferencia en Redes Neuronales ante el Derecho
To approach the phenomenon of bias generated through neural networks, either in their training or in the design of their objective function, and to analyze some of its possible legal implications.
Origin: Latin America and the Caribbean
Language: Spanish
Type: Guide
Creator: Carlos Amunátegui Perelló 1* Raúl Madrid for CeTyS

Ethical OS Toolkit
For visualizing and anticipating future risk of technology products, acknowledging that once technology is released and reaches scale it may be used for purposes beyond the original intention.
Origin: North America
Language: English
Type: Guide or Manual
Creator: The Institute for the Future, Omidyar Network‘s Tech and Society Solutions Lab