An algorithm that reads your social media feed, gets a fix on your digital personality and analyses your
online interactions with friends and family may now be able to determine if you get a loan.

The Law Ministry selected six firms last month (including Minterest) to pilot new business models for moneylending while also temporarily lifting a 2012 moratorium on new licences needed to operate in the sector.

The idea of using personal data to forecast a person’s likelihood of default is being explored by some
licensed moneylending firms here in what is an industry first.

The Law Ministry selected six firms last month to pilot new business models for moneylending while also temporarily lifting a 2012 moratorium on new licences needed to operate in the sector.

Its moves come as personal loans are on the rise. Singaporeans accumulated nearly $60 billion of debt,
excluding housing, credit cards and car loans, in the third quarter of last year.

The key factor for moneylenders is the credit score, which is used to discern a person’s financial reliability based on his credit track record and income levels.

But determining risk of non-repayment, even with the already strong credit bureau that Singapore has, is not an exact science, say those participating in the trial.

Credit reports may not necessarily reflect a borrower’s immediate financial situation, such as when they
have lost a job recently, said our co-founder Ronnie Chia.

Psychometric analysis, which feeds on social media and other online data from borrowers, allows lenders to determine the character and personality traits of borrowers.

These are important indications of the willingness to repay loans, said Mr Chia.

Mr Jonathan Chong, IFS Capital’s vice-president of business planning and analysis, said lenders will be able to assess the borrower better and faster by considering these additional data points.

“With permission, we can overlay the digital footprint of a user, including social and professional
connections, with the personal and financial data to help us form a more holistic picture of the (borrower),” said Mr Chong.

A more positive credit score means a borrower will get a more favourable interest rate that reflects his risk profile, said Mr Chia.

But there are dangers in giving the computer full autonomy in deciding whether a person is likely to go into delinquency, given that the accuracy of AI-driven risk assessments is sometimes questionable.

United States-based non-profit organisation ProPublica found how risk scores produced by artificial
intelligence were unable to accurately tell if a defendant is likely to commit a future crime, even though
these scores are already used in US courtrooms to guide judges during sentencing.

Its 2016 study found that the algorithm, which was developed by a private company, was “remarkably
unreliable” – only one in five people flagged as likely to commit violent crime went on to do so.

It also found that the system unfairly pegged African-American defendants to be at a higher risk of
committing crime than white Americans.

And even when accurate, the use of credit scores to determine risk and reward can be controversial.

China’s newly deployed social credit system, for instance, has been decried as Orwellian by critics for
penalising everyday behaviour.

Chinese citizens have been barred from buying flight and train tickets or purchasing property in recent
months due to a low score from loan defaults or other activities deemed to be anti-social.

These dangers are why there is a need for the Law Ministry’s pilot, so that the personal lending industry
can find out what safeguards are necessary as it heads towards the adoption of AI and big data analytics,
said Mr Edmund Sim, founder of fintech start-up Credit Culture.

However, even detractors of AI cannot deny the superhuman efficiency of computer algorithms, big data
and machine learning employed in various industries today.

Humans can get it wrong even without AI, and they do so while incurring higher costs that will eventually show up on the borrower’s bill.

In a demonstration at his Craig Road office, Mr Sim showed The Straits Times how its credit scoring engine analyses and scores applicants for their creditworthiness almost instantly.

He declined to give specifics on the algorithms used as they are proprietary to the firm.

While accuracy is a concern, machine beats man on one key element – algorithms apply their standards
universally and eliminate any form of bias associated with human intervention, said Mr Sim.

“Technology can benefit consumers by making things simpler, cheaper and more transparent, allowing
them to make better informed decisions,” he noted.

“The question now is the extent to which technology can be used in personal lending.”