Responsible AI, FEAT & Veritas - Dr. Li Xuchun AI Development Office, FinTech & Innovation Group, MAS - Telecom Paris
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Agenda 1 Responsible Use of AI and DA, FEAT 2 Veritas Background 3 Veritas Phase One 4 Veritas Phase Two Plan 2
Why are we concerned with the responsible use of AI and DA? “What all of us have to do is to make sure we are using AI in a way that is for the benefit of humanity, not to the detriment of humanity.” - Tim Cook, CEO, Apple
FEAT Principles Fairness Ethics With data, there are Are there are certain infinite possibilities of lines that should not what we can do. be crossed? But, should we? Accountability Transparency Do we understand To what extent how data-driven should explanations decisions are made? be given for data- Who bears the driven decisions? responsibility and risks of these decisions? 5
Examples of the FEAT Principles Fairness Ethics Accountability Transparency • Ensure that • Use of AIDA is • Internal •Disclosure of individuals or aligned with accountability AIDA use to data groups are not firm’s ethical to the firm’s subjects systemically standards, values stakeholders disadvantaged and codes of (management, •Provision of clear through AIDA- conduct Board) explanations on driven decisions data used to unless these • AIDA-driven • External make AIDA- decisions can decisions held to accountability driven decisions be justified same ethical to data subjects and how it standards as impacts them • Regular review human-driven of models and decisions •Provision of clear AIDA-driven explanations of decisions to how AIDA-driven minimize bias decisions impact and ensure data subjects models behave as designed 6
Global AI Governance Landscape Nov 2018 Feb 2019 May 2019 Nov 2019 Hong Kong Apr 2021 Hong Kong Singapore Japan OECD HKMA High-level The State of Ethical AI in Hong Kong MAS FEAT Social OECD Principles on Artificial Principles Principles Principles Intelligence Apr 2021 European Commission of Human- on AI Proposal for a Regulation laying Centric AI Nov 2019 Australia down harmonised rules on artificial AI Ethics Framework intelligence Apr 2021 U.S. Federal Trade Commission guidance on AI Mar 2021 Korea Financial Services Commission announces a new policy framework on insurance business Jan 2019 Apr 2019 EU Jun 2019 China Singapore Ethics Governance Mar 2021 The five largest federal IMDA Model guidelines Principles for the financial regulators in the United AI for New Generation States released a request for Governance trustworthy Artificial information on how banks use AI Framework AI Intelligence 8
Veritas Challenges Objective Practical implementation of FEAT Creating a standardized and modular principles: Fairness, Ethics, implementation methodology of FEAT Accountability and Transparency. principles and programming open source code Consortium Impact Development of responsible and Formation of industry consortium inclusive financial services by using to develop Veritas to validate AI AI and Machine Learning in a models against FEAT responsible way 9
Scope of Veritas Phase One in 2020 Veritas was announced at 2019 Singapore FinTech Festival as part of Singapore National AI Strategy by DPM. Veritas phase one: focus on fairness principles of FEAT. Veritas phase one: two banking use cases were selected which are credit risk scoring and customer marketing. 11
Phase One Deliverables : Whitepapers and Python Open Source Code Whitepaper to document the generic Whitepaper of use case studies for credit methodology to assess the fairness for AIDA risk scoring and customer marketing. system. Python code for Credit Risk Python code for Customer use case Marketing use case 12
A Comprehensive Five-part Methodology to Assess Alignment with FEAT Fairness Principles A Describe system objectives and context B Examine data and models for unintended bias AIDA System Developer C Measure disadvantage AIDA System Deployer AIDA System Assessor D Justify the use of personal attributes E Examine system monitoring 13
Veritas Fairness Assessment Flow Decide on Complete Address Define scope of Adapt Gather required measures and analysis and feedback and assessment assessment information metrics review report on results Choices of Self-assessment Fairness All information Fairness measure and report and External assessment Part required to assessment Part metrics and feedback from reporting A questions complete B-E questions justifications for internal assessment those choices stakeholders 14
Veritas Programming Open Source Code 15
4. Veritas Phase Two 16
Plan of the Veritas Phase Two in 2021 SCOPE DELIVERABLES Lead Members Use Cases 01 CONSOLIDATED WHITEPAPER FAIRNESS Insurance Underwriting Technical Technical Technical Document 1 Document 2 Document 3 Lead Members Use Cases (Fairness) (Ethics & (Transparency) Accountability) ETHICS & • Customer marketing (Banking) ACCOUNTABILITY • Fraud detection (Insurance) 02 TECHNICAL DELIVERABLES Lead Members Use Cases Source Code 1 Source Code 2 TRANSPARENCY • Credit risk scoring (Banking) (Fairness) (Transparency) • Customer marketing (Banking) 17
Veritas Consortium Members in the Phase Two LEAD MEMBERS MEMBERS (FIS) FACILITATORS ADVISORY MEMBERS Drive the Veritas development Provide domain knowledge & Facilitate the whole program Provide expertise & review the of the use cases feedback on the use cases whitepaper 18
Thank you!
You can also read