How PayPal's AI-based lending system works

80
2
How PayPal's AI-based lending system works

It's completely free and we guarantee you'll learn something new every day.

PayPal's system serves as a large-scale transaction sorting system, with your trustworthiness determined based on where your transactions are sorted. This system, using what PayPal calls clustering algorithms, predicts relationships between transactions based on the recipient of the money, the amount sent and the time of the transactions. These predictions guide the system in classifying and clustering certain transactions.

The confidence score is assigned to each transaction within a group based on transaction time and dollar amount, and the transaction type is determined by these classifications. The score is associated with a user's account and recalculated based on further transactions.

The confidence score can be used to make several decisions related to the user, PayPal said. It can help 'indicate the risk of lending money' to certain users, or guide who a user may want to or not want to borrow money from. This score may affect interest rates, determine the amount of credit a user has, or determine the currency that a user can borrow using a peer-to-peer loan.

We have seen a number of security and fraud intelligence patents from PayPal seeking to make its systems more secure, including user-end tools like voice-based biometric authentication and back-end data storage techniques that aim to obscure personal information in photos. With a population of hundreds of millions, the company has a deep-seated interest in fortifying its cybersecurity walls.

A secure platform also is important for gaining the same public trust that traditional financial institutions have, said Chase Norlin, head of cybersecurity workforce firm Transmosis. While fintechs like PayPal have grown in popularity over the past few decades, conventional financial institutions like Bank of America, Wells Fargo and JPMorgan Chase have gained trust mostly because of their technological prowess, he said.

's the world generates more trust than a brand that's just been around for a long time. Norlin said he was confident he would be able to continue working with the team at the end of the year.

While a system like this could bring about potential privacy concerns, Norlin said that utilizing data to keep users from being defrauded and reduce the risks associated with peer-to-peer money transfers could further instill customer trust. This also may solve the issue of bias present in many AI-based lending scenarios, he said. With a system like this, its bias is solely based on how PayPal creates and implements it. If it's based solely on financial transaction-related data as the patent intends, without considering a user's personal information, it could be incredibly egalitarian, he said.

In looking at it from a pure data perspective, there may not be any bias.