The granting of a loan must conform to all kinds of parameters. The more they are valued, the less risky the loan will be, because that way the creditworthiness of the applicant is better known. Thanks to artificial intelligence (AI), algorithms are replacing humans in this kind of analysis because they are capable of making more precise estimates.
That is the mission of Smart Finance , an AI application that uses only algorithms to grant millions of small loans in just a few seconds of analysis.
In this way, the applicant must only allow full access to Smart Finance to collect some of the data stored on his mobile phone, instead of manually entering the amount of money he earns and other scales such as those traditionally provided to evaluate the probability. the loan is repaid.
Analysis data algorithms deep learning (deep learning), a class of more sophisticated machine learning that allows the artificial intelligence learns by itself, is not the most obvious, as the money available to the user or account your history with any non-payment.
For Smart Finance, the important data are those that would not reveal anything important to a human being: how much battery the mobile has left, the speed at which the user enters the date of birth, how often they order take-out food, if any. it took enough time to read the user agreement and other parameters that end up forming a kind of digital fingerprint capable of predicting whether the borrower will pay the requested loan.
In total, Smart Finance helps build a credit rating system based on 1,200 data points related to user behavior . The service then connects potential borrowers with lenders.
However, one of the problems when training machine learning algorithms is that the cause-effect links are blurred. That is, the algorithm may have detected astonishing correlations whose causes are not in sight. For example, the algorithms may have found a correlation of the type: “users with a battery below 12% only repay their loans 43% of the time.”
This strange correlation has been identified by algorithms after diving into large amounts of data, but our human mind is unable to process such amount of information, so we also reveal ourselves unable to explain what the underlying causal relationship is . The best we can hope for is to verify that this correlation exists and is consistent, but not the reason that a person with that battery percentage is likely to breach their contract.
So much so that, in 2017, Smart Finance granted more than two million loans a month with very low default rates, comfortably exceeding the evaluations carried out by traditional banks. Not in vain, the founder of Smart Finance, Ke Jiao , qualifies the metrics that his application contributes about the solvency of the users as the “new standard of beauty”.
Of course, this raises another thorny issue: privacy . These applications dive to unsuspected limits in the data of our daily lives. In fact, many platforms that track smartphone usage have access to data such as location services, phone contact lists, and call logs that can be used to later track and harass delinquent borrowers.
Finally, then, we are facing a more precise system when granting loans, but also at the mercy of correlations that we are not able to understand, as well as much more exposed to our private data being used for illicit purposes.
