Fair play: reporting back from the Veritas Challenge finals

AI doesn’t just create unprecedented opportunities for business. It brings great responsibility too. It’s why there’s such an intense focus on responsible AI across industries. In the financial sector, a new hackathon-style event recently launched in Singapore to fast-track progress in this space.

Called the Global Veritas Challenge 2021, it’s part of the broader Veritas initiative, which aims to provide financial institutions with a verifiable way to incorporate Fairness of the  Fairness, Ethics, Accountability and Transparency (FEAT) Principles into their AI and data analytics (AIDA) solutions. This includes creating a standardised and modular implementation methodology of FEAT principles and open-source code.

Organised by the Monetary Authority of Singapore (MAS), the ASEAN Financial Innovation Network (AFIN) and Accenture, it aims to spur innovation in AIDA solutions for financial institutions (FIs).

A few weeks ago, the 10 finalists in this year’s Veritas Challenge presented their prototypes at a Demo Day. Each prototype had been designed to help FIs operationalise responsible AIDA-driven solutions in their organisations.

Three winners were eventually selected by the panel of judges. On 9 November, they were announced at the Singapore FinTech Festival where they each had the opportunity to presented their solutions to a global audience, along with a cash prize of $50,000 for each team. 

The run-up to the finals

Having asked various industry players to identify key challenges they encountered in implementing the FEAT principles, we narrowed down these challenges to four areas where responsible AIDA solutions are highly relevant: product marketing, risk compliance and fraud monitoring, loan origination and KYC, and credit scoring and profiling (see box-out below).

1. 1. Product marketing: How can FIs and FinTechs promote financial inclusiveness by ensuring that fairness is incorporated into AIDA marketing solutions?
   2. Risk, compliance & fraud monitoring: In an Accenture global survey of risk managers, 58% identified AI as the biggest potential cause of unintended consequences over the next two years. How can FIs ensure that fairness is incorporated into fraud detection AIDA models?
   3. Loan Origination & KYC: According to a World Economic Forum survey, 52% of FIs use AI to service lending customers. How do they ensure fairness is built into underlying processes while providing the unbanked with access to financial services?
   4. Credit scoring/profiling: In the drive to automate credit decision-making, 60% of FIs use new/alternate forms of data and AI models, as well as external data. This risks disadvantaging some groups (e.g. SMEs). How can they reduce disparate treatment and build confidence in their decisions?

Next, we reached out to our global network of FinTechs, FIs and tech providers. The ask for them? Submit proposals for new solutions that could address one of these four key challenges.

On the day

Each team had five minutes to pitch and demonstrate their solution to the judges, followed by seven minutes of Q&A.

Along with Lee Joon Seong, our Applied Intelligence Lead, the panel included heads of data and innovation from some of the world’s largest banks, as well as the Bank of International Settlements (BIS), Singapore’s Personal Data Protection Commission (PDPC) and the Gradient Institute.

They evaluated each of the finalists against a number of criteria. Emphasis was given to how successfully a team integrated the fairness objective into its solution. Overall feasibility, impact and effort were also major factors.

Meet the winners

Let’s take a closer look at the three winners – Cylynyx, TruEra & Demyst, and Visa. Their solutions exemplify the level of innovation in this critical space. First, Cylynyx. Their product provides a way for fairness to be integrated into algorithm development, testing and data handling processes to ensure there’s no unintended bias against groups of individuals, while protecting FIs against potential fraud.

The joint proposal by TruEra & Demyst provides a way for fairness to be incorporated during data sourcing and processing to ensure fair assessment results in ML-based credit scoring. It’s a key breakthrough. Up to now the use of “non-traditional” or alternate data has potentially accentuated the risk of bias.

Visa’s winning proposal tackled the challenge of bias creeping into targeted online marketing using AI. The team has designed a data, model and platform-agnostic fairness toolkit, tailored for FIs, that can screen for and mitigate any unwanted biases in marketing propensity models.

Taken together, these three solutions advanced fairness as a key feature of AIDA in this industry. Like I started by saying, this is absolutely crucial. Increasing use of AIDA enables FIs to deliver better experiences to their customers, as well as transform back-office efficiency. But sustaining and building trust in how AIDA is deployed is a central consideration. These three finalists do just that.

To learn more about Global Veritas Challenge 2021, please check out the post-event summary report here.

We’ll be back again next year for the Global Veritas Challenge 2022, so stay tuned for more updates. Thanks for reading.

END

Disclaimer: This content is provided for general information purposes and is not intended to be used in place of consultation with our professional advisors.

Copyright © 2021 Accenture. All rights reserved. Accenture and its logo are registered trademarks of Accenture