MAS announced successful conclusion of the first phase of the Veritas initiative, which saw the development of the fairness assessment methodology in credit risk scoring and customer marketing. These are the first two use cases to help financial institutions validate the fairness of their Artificial Intelligence and Data Analytics (AIDA) solutions according to the Fairness, Ethics, Accountability, and Transparency (FEAT) principles. The Veritas Consortium, comprising MAS and industry partners, also published a report on the fairness assessment methodology and the open source code of these two use cases. These documents detail a five-part methodology to assess the application of the FEAT fairness principles in the two use cases. The methodology addresses the implementation challenges in the responsible use of AIDA and provides an actionable approach for financial institutions to validate their AIDA solutions.
Veritas, which is a part of National AI Strategy of Singapore, aims to provide financial institutions with a verifiable way to incorporate the FEAT principles into their AIDA solutions. The report expands on the risk-based approach, which the methodology takes, to FEAT fairness assessment. It includes background on the current market practice for AIDA risk management in financial services institutions, guidance for developing scoping and scaling processes to apply the methodology to systems with different risk levels, and guidance on integrating fairness assessment into existing risk management practices. The report defines key terms in the principles, then moves to present the five-part methodology, which includes describing system objectives and context, examining data and models for unintended bias, measuring disadvantage, justifying the use of personal attributes, and examining system monitoring and review.
To help assess the fairness risks of consequential AIDA systems, the report presents a comprehensive 18-question assessment. Questions in the five parts of the methodology are designed to help financial services institutions make their objectives for fairness explicit and provide evidence that the system’s operation is aligned with those objectives. Each part of the methodology also contains a set of considerations to assist AIDA System Owners, Developers, and Assessors in performing and reviewing an assessment. Finally, a series of appendices provide additional technical reference material to complement the methodology (Appendix 1); advice on mitigation strategies and actions to improve AIDA system fairness (Appendix 2); resources for software implementations, fairness measurement, and mitigation (Appendix 3); and references for further reading (Appendix 4).
Accompanying this report are a set of “FEAT Fairness Assessment Case Studies” (Document 2). The document on case studies presents complete and detailed case studies of applying the methodology to hypothetical but realistic systems: one for customer marketing and another for credit scoring. It also provides reflections from two financial services institutions (HSBC for customer marketing and United Overseas Bank Limited for credit scoring) that applied the methodology to their real systems as well as series of use-case specific considerations to assist financial services institutions in completing assessments of customer marketing and credit scoring systems. This development marks a milestone for the Veritas initiative and paves the way for the next phase of work. Phase Two will look into developing the Ethics, Accountability, and Transparency assessment methodology for the two use cases in Phase One. Phase Two will also include use cases for the insurance industry. MAS has also released a list of Phase Two Veritas Consortium members.
Keywords: Asia Pacific, Singapore, Banking, Insurance, Securities, Veritas, Fintech, Artificial Intelligence, Data Analytics, Veritas Consortium, FEAT Principles, MAS
Previous ArticleOSFI Consults on Addressing Climate Risks in Financial Sector
ECB published results of the quarterly lending survey conducted on 143 banks in the euro area.
FED finalized a rule that updates capital planning requirements to reflect the new framework from 2019 that sorts large banks into categories, with requirements that are tailored to the risks of each category.
ESAs published the final draft implementing technical standards on reporting of intra-group transactions and risk concentration of financial conglomerates subject to the supplementary supervision in EU.
EBA published the annual report on asset encumbrance of banks in EU.
FED updated the reporting form and instructions for the FR Y-9C report on consolidated financial statements for holding companies.
EBA issued a consultation paper on the guidelines on monitoring of the threshold and other procedural aspects of the establishment of intermediate EU parent undertakings, or IPUs, as laid down in the Capital Requirements Directive.
EC published Regulation 2021/25 that addresses amendments related to the financial reporting consequences of replacement of the existing interest rate benchmarks with alternative reference rates.
BIS published a bulletin, or a note, that examines the cyber threat landscape in the context of the pandemic and discusses policies to reduce risks to financial stability.
HM Treasury, also known as HMT, has updated the table containing the list of the equivalence decisions that came into effect in UK at the end of the transition period of its withdrawal from EU.
EBA published an erratum for technical package on phase 1 of the reporting framework 3.0.