To avoid bias, fintechs must address diversity from the start

The best intentioned efforts to bridge racial or gender gaps, or to create a more inclusive market for a financial project, will likely fail unless there’s careful attention paid to a lack of diversity at the onset.

Companies trying to bring payment innovation are taking a close look at their development teams in an attempt to reduce unintentional biases that result from a project management strategy that does not take an entire population into consideration.

“There has to be a diverse group of people involved in the programming process. You don’t know what you don’t know,” said Valerie Dahiya, a partner at Perkins Coie in Washington, D.C., who has worked on the firm’s efforts to create diverse teams in blockchain and digital currency projects.

One of the problems that prevent technology projects from reaching a diverse audience is the bias that derives from the original input. For projects that use machine learning or artificial intelligence, that can include incomplete data from limited sources.

Since AI and machine learning build off a foundation of data to inform decisions on marketing, product development, sales and compliance, such projects find themselves limited by the choices made at their outset.

Writing for PaymentsSource, Hani Hagras, chief science officer for Temenos, said AI systems “teach themselves” in a way that reinforces biases. The problem is made worse by “black box” AI systems that do not communicate how decisions are made, in theory making decisions worse over time. An alternative is “white box” or explainable AI, which uses nontechnical language to explain how the software works and how decisions based on AI are made.

The data exists to combat biases. The prevalence of social media — which is more user driven — and the rapid production of data that tracks transactions result in a spread of more information at faster speeds.

Better data analysis combined with more diverse project teams can create a more level field at the onset of a project, Dahiya said.

“Diversity needs to be part of the input that goes into coding,” she said. “That team needs to be diverse.”

Synchrony, which has about 16,000 employees, builds project teams that are divided among tasks such as design, development and business strategy. These tasks each include metrics for representation along race, ethnicity and gender, with the hope of encouraging diversity of thought.

The unconscious biases that people have will perpetuate unless a concerted effort is made to ensure inclusion, according to Brian Doubles, president of Synchrony. “You have to be deliberate about it,” Doubles said.

Synchrony over the past few months began work on a series of credit card projects that include telco-affiliated cards and general- purpose cards, a relatively new market for the Stamford, Conn., company, which General Electric spun off in 2015 and has focused mostly on store-branded cards. Synchrony has additionally expanded into secured cards, which attempt to reach new market segments by helping consumers build credit profiles.

Synchrony, which reports 45% of its staff consists of what it calls “ethnically diverse” employees, formed a committee with Doubles and other executives that’s charged with increasing diversity in its workforce, business partnerships, suppliers and customers. It also made about $5 million in donations to nonprofits and to support organizations that address issues that disproportionately impact the Black community such as the Thurgood Marshall College Fund, National Urban League and Equal Justice Initiative.

Synchrony did not release detailed data on the makeup of project teams, but in addressing bias, Doubles said the company does not use racial data or age as part of its underwriting models.

“This will take more than dollars, so it’s about harnessing the power of our staff to direct real change,” Doubles said, adding the company's investment arm is also making a concerted effort to spot startups owned and operated by a more diverse range of entrepreneurs.