The Bank of England (BoE) published a paper on software validation and artificial intelligence in finance. The use of machine learning and artificial intelligence in finance poses growing risks for software validation to financial institutions, markets, and decision makers, making it a key priority for regulators. This paper discusses accepted software validation practices, highlights challenges to those practices introduced by artificial intelligence and potential solutions, and suggests areas of focus for developers when creating artificial intelligence-based solutions for the finance industry. The paper also discusses how practices may need to evolve to respond to these new challenges and is intended to inform policymakers and governance bodies, while also raising awareness among decision makers in financial institutions.
The paper concludes that the following key points should be borne in mind by financial institutions when developing artificial intelligence/machine learning solutions to support the provision of financial services:
- Machine learning software development is data-driven, making the technology hard to test conventionally, and the challenges are further exacerbated by end-to-end machine learning systems. Software validation may need to move from testing based on requirements to validation based on representative test datasets. These should include corner cases or tail event cases, representing scenarios not catered for by training datasets.
- Machine learning “black box” nature can make it impossible to interpret how decisions are made. Explainability techniques may help attribute which factors are most important in a decision making process, but this may not enable identifying which part of a big machine learning model framework is responsible for any undesirable model behavior. Entanglement can also mean inputs are not independent with complex interdependencies between machine learning components. Decomposing machine learning models into smaller parts to generate decisions can add clarity although such opportunities may diminish as machine learning architectures become more end-to-end.
- The characteristics of training datasets fundamentally influence machine learning model behavior, potentially replicating or amplifying dataset bias. Training datasets must be validated to ensure they are correct and representative, addressing outlier data elements and faulty labels. Datasets used for different purposes such as machine learning training, calibrating machine learning models or checking the accuracy of machine learning models should be free of common biases or flaws to ensure that they are fit for the specific use case that they are applied to. Emergent solutions exist to ensure that datasets are fit for purpose including: repeating data selection in a random way; formally documenting the composition, collection process, recommended uses, and inherent biases of datasets; development of methodologies to detect faulty or skewed datasets; and network graphs to visualize datasets and highlight data relationships graphically.
- Using parallel processing to support machine learning models can result in unintended or inconsistent outputs disruption if the ordering of computational steps and processing takes place out of sequence because of poor overall modeling framework. It is important that the machine learning models, particularly when there are interdependencies among components and different sub-models, have robust controls over the ordering of computation steps.
- Machine learning models are non-deterministic in nature. Some commentators have observed challenges associated with integrating non-deterministic machine learning models with software components that are deterministic/procedural in nature, when, for example, the output of an machine learning model changes qualitatively over time, due to re-calibration, impacting integrated software components.
The paper also provides, for consideration of the policy makers and firms’ governance bodies, a checklist for artificial intelligence software validation. The paper, however, does not focus much on application-specific challenges, which can be considered as a next step where more opinions from subject-matter experts can be incorporated. Similarly, as a next step, one can try to work on more application-specific financial regulations highlighting any gap in existing regulations in a more explicit manner. Machine learning, and artificial intelligence in general, can also help to automate a lot of the existing testing processes and may improve the existing capacity to test software, improving resilience. Therefore, creating the right framework for artificial intelligence software testing could yield wide-reaching benefits, with the appropriate regulatory focus.
Keywords: Europe, UK, Banking, Artificial Intelligence, Machine Learning, Regtech, Software Validation, Modeling Risk, Model Explainability, BoE
The UK authorities have published consultations with respect to the Basel requirements for banks. The Prudential Regulation Authority (PRA) published the consultation paper CP16/22 on rules for the implementation of Basel 3.1 standards.
The three European Supervisory Authorities (ESAs) issued a letter to inform about delay in the Sustainable Finance Disclosure Regulation (SFDR) mandate, along with a Call for Evidence on greenwashing practices.
The Financial Stability Board (FSB) and the Network for Greening the Financial System (NGFS) published a joint report that outlines the initial findings from climate scenario analyses undertaken by financial authorities to assess climate-related financial risks.
The Financial Stability Board (FSB) published a letter intended for the G20 leaders, highlighting the work that it will undertake under the Indian G20 Presidency in 2023 to strengthen resilience of the financial system.
The International Sustainability Standards Board (ISSB) of the IFRS Foundations made several announcements at COP27 and with respect to its work on the sustainability standards.
The International Organization for Securities Commissions (IOSCO), at COP27, outlined the regulatory priorities for sustainability disclosures, mitigation of greenwashing, and promotion of integrity in carbon markets.
The European Banking Authority (EBA) issued a statement in the context of COP27, clarified the operationalization of intermediate EU parent undertakings (IPUs) of third-country groups
The European Union has finalized and published, in the Official Journal of the European Union, a set of 13 Delegated and Implementing Regulations applicable to the European crowdfunding service providers.
The Office of the Superintendent of Financial Institutions (OSFI) published an annual report on its activities, a report on forward-looking work.
The Australian Prudential Regulation Authority (APRA) finalized amendments to the capital framework, announced a review of the prudential framework for groups.