The bank, which makes much of its money in Asia, has collaborated with US-based data scientists to improve the quality of machine learning decision-making across its services.

Machine learning has become the de facto tool for trawling masses of data to identify patterns on which many risk management decisions are made. But these complex automated algorithms are not bullet proof when it comes to deciding whether those decisions are reached openly and fairly.

Truera, based in Redwood City, California, worked with Standard Chartered's retail analytics, risk and technology teams to look at the lender's credit decision-making algorithm which probes both traditional and alternative data (with customers’ consent) to see where biases could be creeping in. Truera's developers say the platform essentially adds a whole layer of explainability to AI, based on causality and cooperative game theory as well fairness and bias.

The group was able to look for correlations between what appeared to be impartial variables, concerning race or gender for example, but which could lead to unfair decisions being made. The makers say that the tool also detects patterns of how data is distributed which can also result in poor predictive outcomes over time.

Truera's president and co-founder Anupam Datta, on leave as professor and director of the Accountable Systems Lab at Carnegie Mellon University, said that customers have been using the platform more lately to understand anomalies thrown up by COVID-19. Research for the technology began in 2014, when a team at the university discovered gender bias in online advertising but lacked the tools to measure its effects. Last year, the startup raised $5.1 million led by Greylock Partners in a space that is gaining momentum as corporates and other institutions tackle issues of bias across their operations.

As AI and ML are becoming more relied on, regulators too are getting into the weeds of how financial institutions are making AI-based decisions so that they can justifiably explain the outcomes these data systems recommend.

StanChart is one of several banks operating in Asia signed up to the Veritas consortium in Singapore. Veritas was launched by the Monetary Authority of Singapore along with sector incumbents a couple of years ago to strengthen internal governance of AI and data management. The framework allows financial institutions to evaluate how they are using models based on fairness, ethics, accountability and transparency, otherwise known as meeting the FEAT principles. Microsoft, StanChart, Deutsche Bank, Visa and Linklaters are among those providing early feedback to MAS. StanChart is also part of the Artificial Intelligence Public-Private Forum in the UK which is soliciting similar best practice insights.

“New developments in analytical technology and expanding usage of data require us to fundamentally rethink how we demonstrate ongoing adherence to our pillars and tackle the issue of unjust bias,” Sam Kumar, global head of analytics and data management at Standard Chartered, said.

Its partnership with Truera is intended to "give confidence to customers and regulators" in the fairness of outcomes, the bank added.