Coding the right values into artificial intelligence

Artificial Intelligence — technology with the power to absorb billions of pieces of data, analyze its contents, then use that information to learn and make decisions in real time — is here. No longer simply the plot from futuristic Hollywood action movies or sci-fi novels, AI is now powering a future filled with potential.

Is the financial services industry ready to deal with this kind of disruption? What happens when machines learn faster than humans, when automation drives service delivery, and when software makes decisions? How does the financial services industry ensure we’ve coded the right values into the technology in order to avoid unintended consequences in its application?

Earlier this summer, TD Bank Group brought together some of the best thinkers in the AI field — academics, bank technologists, fintechs, global consultants, public sector leaders — and asked them to work with us to identify critical areas of focus for the financial services industry. The group analyzed different possible scenarios that could impact our industry, our customers and society at large and worked backwards to pinpoint the critical decisions that could inform the outcomes. We identified three fundamental and interconnected areas of focus that should be tested and regularly reviewed and refined — by humans.

The first is “explainability.” If people are to trust the results from AI insights, they have to be able to understand and explain the outcomes. This sounds basic but it’s not. The human brain simply cannot compute at the speed and scale of AI and we must therefore make sure that as the software learns, we understand how and why it is arriving at conclusions. Think of it this way: if you are approved (or not) for a loan, the banker will walk you through the rationale for the decision. AI must allow us to do the same. It needs to augment human judgment and decisions in a way that’s transparent and explainable.

A second focus is “negative bias.” AI must not develop bad habits or use data to pre-judge outcomes without taking unique, human situations into account. If programmed or used incorrectly, AI can learn the wrong lessons and start categorizing people and data into buckets that unfairly disadvantage certain groups or demographics.

The third area we identified was “diversity and inclusion.” From management to computer scientists, the diversity of the teams that write the code that forms the backbone of this technology is paramount. If you staff an organization without

WE’VE TAKEN IMPORTANT STEPS BUT THERE IS MUCH MORE

TO DO.

diverse representation, it may impact that company or organization’s ability to serve its customers. The same principles are true of AI. If the code is created and its performance overseen by people who do not represent those served, the outcomes that result could unfairly disadvantage individuals or even entire groups.

As an industry, we need rigorous processes to preserve the trust that repeated polls indicate Canadians have in their banks. With AI, as with any business decision at the bank, preserving that trust remains central to our innovation strategy. From cybersecurity to customer education, financial institutions invest heavily to protect and secure the firms’ and customers’ data and privacy. As the financial services industry embraces AI, it’s important to continue reflecting on and analyzing risks and potential customer impacts.

We’ve taken important steps but there is much more to do. At TD, we support explainability and test for bias in our machine-learning models. And we are continuously testing and verifying the results they produce, with human oversight at every step.

Ultimately, banking is a people business. The financial services industry doesn’t deliver shiny products or cool gadgets. We provide services that enrich lives and help build financial confidence. Our overriding focus must be that people always stay in the driver’s seat. AI is not really about delegating work to robots. It’s about helping people achieve more precise outcomes more efficiently and more quickly.

The financial services industry is embracing AI to power the future of customer experiences. As an industry, we have to make clear to our customers that we will do this responsibly and then demonstrate how we are living up to that promise.