This lender is using facial-expression software to detect fraud
Can you tell if your customers are lying? A Chinese lender believes it can use face-tracking technology to spot dishonest customers.
Ping An, a Chinese financial services conglomerate, has developed technology that tracks micro-expressions in order to detect signs that customers may be untruthful. Chinese banks, worried that some customers may be lying about the uses of loans, have begun implementing the technology commercially.
Ping An’s technology can identify 54 involuntary micro-expressions that people often make before they have a chance to exert control over their facial movements, according to a report by The Financial Times. These micro-expressions, often there and gone in 1/15 of a second, are difficult for humans to detect.
Chinese banks are among the first to use the micro-expression technology commercially, according to The Financial Times. Western companies have been hesitant to adopt it because of concerns about the ethics of using it – and the tech’s basic reliability.
But Lee Yuansiong, a deputy chief executive at Ping An, insisted that the face-tracking tech worked.
“We use micro-expression recognition technology to review loan applications. It captures subtle changes in customers’ facial expressions, which help to identify and warn against fraud risks,” he told The Financial Times. “We’ve reduced credit losses by 60% using this technology. It is more accurate than other approaches to fraud detection.”
Others, however, question the technology. Husayn Kassai, co-founder of facial-recognition company Onfido, told The Financial Times that academic studies into the efficacy of micro-expression tracking had “not always yielded consistent results.”
“Besides, even if the feasibility is proven through research, there are serious ethical questions around such a case,” he said. “For instance, such a system would probably need to be used without the user’s awareness in order to capture natural micro-expressions – which raises questions around a user’s informed consent.”