It’s a first: OCR Labs Global claims it is the first company to independently verify that its identification verification software has been certified bias-free.
In a future where we can expect to identify ourselves to access all kinds of products and services — not just online but in-person — news of zero-bias identification verification methods is encouraging.
The UK-based firm announced on Jan. 31 the results of an evaluation performed by BixeLab. In the first test attempt, this independent biometric testing company found that OCR Labs’ facial liveliness detection software has zero demographic bias and no error rates.
“The time has come to refocus efforts on achieving inclusivity to prepare for the future when more people than ever will use identity solutions for everything, even unlocking our car with our face,” Russ Cohn, general manager international for OCR Labs Global, said in a media release. “With our Zero Bias AI, education, and strategic partnerships … we are helping to facilitate that future.”
The BixeLab evaluation was performed on candidates aged 18-70 who are male, female, and transgender from eight ethnic groupings.
Solving biometric bias issues ‘essential’ to business
OCR Labs, in the release, references research that shows women and ethnic minority groups “are most at risk” from the failure to build inclusive identity verification procedures.
In a white paper on the firm’s efforts developing the software, authors, while detailing dataset bias, pointed to a study from MIT that showed facial recognition error rates for white men were less than one percent, while for dark-skinned women, the error rates were one in three.
Solving the problem of biometric bias in identification verification is not only moral and important to meet increasing government standards, argue the white paper’s authors, it’s also ‘essential’ for businesses trying to reach as many potential customers as possible, regardless of race, ethnicity, sex, gender identity, age, national origin or disability.
“We believe that by removing barriers that lead to exclusion, digital ID systems can ensure that as many people as possible have access to life-enhancing products and services, no matter their background,” Cohn said.
The white paper details the software’s development process, which considered biases in systems (such as mobile phone operating systems), datasets, language, and manual screening processes. It describes how many companies use the software in their work.
“Using our technology, Bloom Money helps diaspora communities across Europe manage their finances seamlessly through their app, overcoming data bias,” Cohn said. “Likewise, Reed Screening is actively tackling manual screening bias using our software to make hiring faster, fairer, and safer.
“And Admiral Money, a subsidiary of Admiral Group PLC, is utilizing our solution and research cataloging 110 languages, 32 typesets, and 16,000-plus identity documents to facilitate access to home improvement loans, debt consolidation, and car financing products to traditionally underserved groups that speak, read and write multiple languages.”
Craig is a freelance writer and editor. He has toiled in various positions for various newspapers in Western Canada, including the Edmonton Journal and the Calgary Herald.
When he’s not busy fixing his home, you can find him experimenting with his slow cooker, finding the right grind (and coffee bean) for his AeroPress, reading fiction and non-fiction, mulling over director Ingmar Bergman’s works, and practicing his backward crossovers (both sides!) while ice skating.