人工智能伦理荟萃

课程、学习资料

• Simple Introduction to ML and DL: McKinsey Executive guide to AI
• Compilation of Academic Syllabi for Ethics and Technology: Medium Post / Google Doc
• Human-AI Interaction Course CMU, Fall 2018
• ACL Web Ethics in NLP Course
• AI Ethics for Middle Schoolers (Blakeley Payne)

相关报告和论文

• Obama White House AI Report

• Human Rights:

• Human Rights and AI (Harvard paper): setting an agenda for the topic

• Evan Selinger a Human Rights framing for AI Ethics: post / thread

• Governance with teeth (Vidushi’s Article19 paper)

• Rights aren’t the best way to define ethics or policy (Pranesh’s thread)

• Manifesto on Algorithmic Humanitarianism

• AI and Human Rights by Harvard

• Governance:

• Governing AI (Data&Society)

• Berkmen Klein 2019 Assembly Projects (Focus: AI Ethics & Governance)

• Street level algorithms (Ali Alkhatib): paper | CADE seminar

• Google Perspective on AI Governance

• Comparative legal study on privacy and personal data protection for robots equipped with AI: looking at functional and technological aspects.

• The Principles and Limits of Algorithm-in-the-Loop Decision Making (paper)

• Disconnects b/w tech and law: exploring “Is there a disconnect between how law operates and how algorithmic decision-making systems work? Process vs outcome? Vagueness by design vs precision?”

• Human Judgement in Algorithmic Loops: Individual Justice and Automated Decision-Making (Case for Individual judgement vs discretion and its limits).

• AI Now Reports:

• 2018 Impact Assessment

• Gender Race and Power

• 2019 Symposium Talks

• 2019 Year in Review

• AI Ethics in India:

• AI in India (CIS compendium)

• AI Policy in India

• AI Ethics in India (Founding Fuel article)

• Establishing “Responsibility” in AI: Artificial Intelligence, Responsibility Attribution, and a Relational Justification of Explainability (paper)

• Kate Crawford’s NeurIPS2017 Talk on Bias in AI/ML

• Google ML Fairness Module

• 21 Definitions of Fairness (Arvind Narayanan)

• Unfairness By Algorithm (Future of Privacy Forum Unfairness/Harms info-graph)

• Tutorial on Algorithmic Bias Detection and Fairness (Francesco Bonchi)

• A Framework for Understanding Unintended Consequences of Machine Learning (Harini Suresh’s paper)

• Blog on how bias creeps into big data analysis (nice plots)

• "Getting Specific about Algorithmic Bias" Rachel’s talk at PyBay 2019 (and thread)

• Bias in NLP:

• Annotator bias in NLP datasets (paper)

• Man is to Computer Programmer as Woman is to Homemaker? (paper)

• Bias in word embeddings: FRAGE (paper)

• Bias in NLP: Let's Talk About Race: Identity, Chatbots, and AI

• Cool project with code to evaluate bias in Harry Potter text (blog)

• Bias in Biometrics/PII categories:

• PEW report on predicting gender using DL: Essay / Interactive

• Bias in ASR (paper)

• Fairness Research:

• Fairness in representation (paper): quantifying stereotyping as a representational harm: Formalizes two notions of representational harm caused by "stereotyping" in machine learning and suggests ways to mitigate them.

• Frontiers of fairness in ML (paper)

• Are We Safe Yet? The Limitations of Distributional Features for Fake News Detection. (paper)

• Fairer machine learning in the real world: Mitigating discrimination without collecting sensitive data. (paper)

• On Formalizing Fairness in Prediction with Machine Learning (fatml doc)

• Fairness through Causal Reasoning (paper)

• Recourse, Explainability, Transparency, Interpretability:

• Actionable Recourse in Linear Classifications (paper)

• FairTest: Discovering Unwarranted Associations in Data-Driven Apps (paper)

• Accelerating Human in Loop ML (paper on challenges & opportunities)

• Explainability of ML predictions: Why should I trust you (paper)

• FAT* 2019:

• Reflections on FAT 2019 (William Isaac)

• FAT* 2019 Papers

• Issues with Predictive Policing (Suresh V)

• On centering fairness on the harmed vs the agent/tech (Suresh V)

• Issues with Pretrained AI Models (Bias, Governance etc) (paper)

• The Politics of Images in Machine Learning Training Sets: Excavating AI

• Open AI GPT2 Ethics Debate:

• Paper/update with bias discussion

• GPT2 Model Card

• Post on final model release and learnings

• Facebook’s suicide prevention algorithm (paper)

• Predicting Mental Health from Social Media (paper)

• Carbon footprint of AI (paper)

• AI & Climate Action - google doc by John Chavens to bring together on this topic. AIforGood:

• How to approach #aiforgood (Erini’s blog)

• Mechanism Design (#aiforgood paper): Describes the Mechanism Design for Social Good (MD4SG) research agenda, which involves using insights from algorithms, optimization, and mechanism design to improve access to opportunity.

• “Dagsthul Declaration for #aiforgood", similar to Toronto and Montreal Declarations.

• Google #aiforgood grantees learnings report

• AI and Conservation (Book)

• Tackling Climate Change with AI (paper)

• Examples of #aiforgood applications (Jack Clark’s tweet thread)

• Google India #aiforgood workshop videos

• Using AI for Human Rights Monitoring: Element.ai post | paper

• CIS workshop: Perspectives on #AIforGood

书籍

• Weapons of Math Destruction - Cathy O’Neil
• Artificial Unintelligence - Meredith Broussard
• Algorithms of Oppression - Safia Umoja Noble
• Race after technology - Ruja Benjamin
• Book on the Morality of Machines
• Studying those who study us
• Shannon Vallor’s Virtue Ethics Book
• DJ Patil’s Ethics & Data Science (free e-book)
• Value Sensitive Design
• Meredith Broussard’s essential T&S reading list

数据集

• DEPA (indiastack blog post)
• Blog on Benefits of DEPA and Account Aggregator
• Justice AP Shah report on Privacy
• Srikrishna Committee Final Report on Data Privacy Law
• Rahul Matthan on Data Protection FW (and consent intermediaries)
• Lawyers submission to SK committee

相关会议

• ACM FAT*
• AIES
• AAAI 2020
• World Summit AI
• NeurIPS
• ICML

附加链接和资源

• Rachel’s ethics resources listing
• Rachel’s thread on Humanities readings for STEM folks
• Eirini’s Critical Reading List for Responsible AI
• Data 4 Black Lives Conference Notes
• Silicon Valley, AI and the Military Reading Materials
• Chat on good AI Ethics research
• Companies building AI Ethics tools (eg Arthur AI)

播客

• Fei-Fei Li Podcast on Human Centered AI
• Shannon Vallor Podcast on Virtue Ethics
• Joanna Bryson Podcast on Human in the Loop
• DJ Patil Interview on AI Ethics

组织和社区

• Berkman Klein (Harvard & MIT Center)
• Center for Applied Data Ethics (USF)
• Human Centered AI (Stanford)
• AI Now Institute (NYU)
• IEEE (John Chavens)
• World Economic Forum 4th Industrial Revolution Center
• Data and Society
• Montreal AI Ethics Institute
• Partnership on AI
• Future of Life Institute
• Alan Turing Institute
• The Institute for Ethical AI & Machine Learning
• Leverhulme Centre for the Future of Intelligence
• Ethical Resolve
• SF AI Meetup (Mia Dand)

相关资料

• Intro to Data Ethics: SCU resource
• Privacy and Liberty in the age of always-on, always-connected world. (paper)
• Ethics Manifesto for Cybersecurity
• Illusion of user choice (blog post)
• Facebook and algorithmic data profiling (tweet thread)
• Links to papers on politics of urban data (tweet thread)
• Lean Data Practice (HasGeek & Mozilla talks)
• Security Planner (Guide for online safety by Zeynep)
• Infosec Resources List
• Digital Safety Practice Guidelines for India (Karana post)
• How email tracking works (freedomtotinker blog post)
• Threat modeling experts (useful to model AI disclosure policies?)
• EFF report on cell-site simulators or StingRays exploitation
• Big Data for International Development
• Digital ID stages (Engine Room report)
• Omidyar stand on data and privacy
• Omidyar report on Civic Tech in India
• Models for Platform Governance (CIGI reports)
• The case for “Public Interest Technologists”

实践和工具

总结

• The “global” landscape of AI ethics guidelines (Nature paper)
• Summary Paper: Ethics of Algorithms (Mapping the debate)
• Linking AI Principles (paper | blog)
• “Global” ethics principles (June 2019)
• BKC Harvard Viz on Principled AI
• Summary of 88 Global Ethics Guidelines
• IEEE Ethically Aligned Design

原则

• European AI Ethics Principles
• Google AI Principles (update after 6 months)
• Unified framework for 5 AI principles (MIT paper)
• 8 Principles for Responsible AI (itechlaw blog)
• East Asian Perspectives on AI Ethics

实践、框架和清单

• Ethical by Design (Ethics Best Practices for NLP paper)
• Ethics Best Practices (Stanford AI blog post)
• AI4People: Ethical Framework (Virginia Dignum)
• Canada Protocol for AI for mental health (checklist)
• Norms for responsible publishing in AI research: Lawfare Blog | PAI Blog

当前方法评论相关资料

• Ethics of AI Ethics (paper)
• Key Gaps in AI Ethics (Microsoft paper, July 2019)
• A Practice-Based Framework for Improving Critical Data Studies and Data Science
• Ethical AI Issues, CIS perspective (blog)
• Roles & Limits of AI Ethics Principles (paper)
• AI Ethics: Too Principled To Fail (paper)
• Better, Nicer, Clearer, Fairer: A Critical Assessment of the Movement for Ethical Artificial Intelligence and Machine Learning (paper)
• 7 Traps of AI Ethics (blog)

AI伦理实践

• Microsoft AI Ethics Practice
• Salesforce AI Ethics Practice
• Google Responsible AI Practice
• From Principles to Practice (Salesforce Workshop)
• Sam Charrington’s Thread on Operationalizing Ethics
• TwiMLcon panel on Operationalizing Ethics
• Axon Ethics Board Report
• Putting Ethical AI Into Practice (blog post)

道德实践工具

• Datasheets for Datasets (paper): Datasheets includes sections with detailed questions on: Motivation, Composition, Collection Process, Preprocessing (Cleaning/Labeling), Uses, Distribution, and Maintenance. This helps team do pre-screening to reduce/mitigate bias.
• Model Cards for Model Reporting (paper): Provides a card to be filled in: Model details, Intended Use, Factors considered, Metrics, Evaluation Data, Training Data, Quantitative Analysis, Ethical Considerations, Caveats and Recommendations
• AI Blindspot Cards (Berkman Klein Assembly Project): A discovery process for spotting unconscious biases and structural inequalities in AI systems through their planning, building deploying stages.
• Markkula Center Resources (Santa Clara University): Toolkit | Slides
• OpenRoboEthics AI Toolkit (Foresight into AI Ethics)

多主题相关资料

• The Ethics of Information – Luciano Floridi
• Technical Debt to Intellectual Debt (blog post)
• Why we all need to care about Data Ethics (Article)
• Understanding Algorithms as Culture (STS paper)
• Understanding Opacity of ML algorithms (STS paper)
• 2011 Paper on Ethics of AI (Nick Bostrom)
• CDEI AI Ethics Snapshot Papers
• Proceedings of the 2019 AAAI/ACM Conference on AI, Ethics, and Society
• BKC Harvard 2 yr Summary: simultaneous work on 5 fronts: 1) disinfo/propaganda and AI 2) policymaking for justice 3) AI governance (beyond US) 4) engaging tech building engines 5) AI ethics education
• Tech Industry Shaping AI Governance Rules (Wired article)
• Algorithmic Colonialism by Rumman
• Algorithmic Colonization of Africa
• History of CS + Ethics study tweet thread
• Moral character of cryptographic work
• PEW report: Half of US ok with Facial recog AI
• AutoML: Intro to AutoML | AutoML Hype

相关技术和主题

• Case studies for Software Engineering Ethics (not AI specific)

• Ethics of IT (plato stanford paper)

• Cynefin FW: Decision making framework for societal challenges

• Citation Guidelines to Decolonize Tech (D&S post)

• Examples of On-ground Activism via Social Media (Rumman’s thread)

• People’s agency over tech (relevant links): Lilly's reading lists / Lilly Irani's tweet thread

• Mozilla Internet Health Report 2019

• Upenn Course: Introduction to Smart Cities (syllabus)

• The philosopher redefines equality (long post on the basis of a free and fair society)

• On Predatory financial inclusion: Kenyan financial lending apps loop of slavery

• On finance innovation as distraction from inequality + tech solutionism (stanford law)

• Ethical Design resources

• Ethical Design Pyramid: based on Human rights, efforts & experiences