Artificial intelligence, or AI, in the movies is all about sentient supercomputers and killer robots. But in real life, AI is nothing of the sort (not yet, anyway). Instead, AI algorithms quietly operate in the background, influencing what appears on your Facebook newsfeed, the best route to take on Google Maps, and perhaps soon – whether you qualify for a home loan or not.
This is not brand new. AI and machine learning have been used in credit scoring models for a few years already. But it has not yet become widespread – nor has it fully reached Singapore’s shores. But, with MAS announcing in May that the first phase of its Veritas initiative (to promote responsible AI adoption for financial institutions) has commenced, perhaps it’s time to take a closer look at this emerging trend.
AI in Credit Scoring Models – What’s It All About?
A credit scoring model is basically how banks judge your creditworthiness. Your credit score is one traditional key factor (which is why it’s so important to keep a good score). Others include your income and the size of the loan you are applying for.
But are these data points enough? Many financial institutions don’t think so – especially in the age of Big Data. The purpose of these AI credit scoring models is to gather many more data points about an individual (even seemingly unrelated ones) and use them to create a model that can better predict an applicant’s likelihood of repaying the loan.
A lot of this data is referred to as “alternative data”. This can include things like information scraped from your social media accounts, your utility payment history, and your credit card purchase history. The more data an AI algorithm can process, the more it can “learn” and theoretically become more precise in estimating a borrower’s credit risk.
This leads to two key advantages.
2 Advantages of AI Credit Scoring Models
Yes, the amount and type of data that can possibly go into these models can appear quite dystopian. But the available evidence demonstrates that it is actually more effective at predicting a borrower’s credit risk – it’s first advantage.
- A working paper by the Bank of International Settlements concluded that the machine learning model was better able to predict losses and defaults than traditional models during times of economic distress.
- Research by China Agricultural University found that social network data was meaningfully correlated with loan default rates.
Now, keep in mind that both studies were done using data from Chinese fintech platforms. This means that these were not home loans, but rather unsecured personal loans. That said, there is no reason why this effectiveness should not extend to all types of loans – including home loans.
The second advantage is greater inclusion. AI credit scoring models are not “stricter” or “looser” than traditional ones – their aim is to simply be “better”. And this means that there may be cases where applicants that may not qualify under the traditional model could qualify under the AI-powered one. This has the potential to boost financial inclusion, especially for those who may face rejection under the traditional model because of insufficient credit history.
But there is a flip side to the coin.
2 Key Risks of AI Credit Scoring Models
Despite these advantages, AI credit scoring models are not without risks. One is that the algorithm may become a “black box”, meaning we have little insight on how the algorithm is making its decisions. This could mean that we would have no idea why the algorithm might have rejected a certain applicant. This is obviously unacceptable when it comes to something as crucial as home loans, where guaranteeing fairness is paramount.
The second key risk is that AI algorithms might inadvertently exacerbate existing biases. As the saying goes “garbage in, garbage out”. If the algorithm is being “fed” data that already reflects certain biases (which might have been the result of historical systemic discrimination), then its decisions will continue to reflect and even enhance said biases – leading to widespread unfairness.
On the surface, these risks don’t appear to be too serious. But imagine if they were amplified throughout every single loan decision in the future. Given how important obtaining credit is to our society, biased algorithms could lead to greater wealth inequalities and other societal issues.
So, how can we mitigate such risks?
How Singapore is Being Proactive in Mitigating Such Risks
There are several technical solutions available to ensure AI credit decisions are as fair as possible. The Harvard Business Review, for example, offers three solutions:
- Removing bias from the data before the model is built
- Explicitly choosing goals for the AI linked to fairness metrics
- Using an opposing AI algorithm to as a “fairness checker”
But the specific solutions themselves are not that important for us – the consumers. What matters is that these solutions, or similar ones, are put in place before AI credit scoring models become predominant. Prevention is better than cure.
Fortunately, in that respect, the Singaporean government has been proactive. As we mentioned in the beginning of this article, MAS is moving forward with the Veritas initiative to promote a responsible adoption of AI by financial institutions.
The goal is for any adoption to adhere to the principles of FEAT – Fairness, Ethics, Accountability, and Transparency. A key point is that individuals or groups cannot be systematically disadvantaged through AI-driven decisions, unless such decisions can be justified. The models must also be regularly reviewed and validated for accuracy.
AI Credit Scoring is the Likely Future (But Don’t Be Worried!)
Currently, most of the AI credit scoring models are limited to unsecured personal loans given out by fintech platforms. But that is changing. For instance, Equifax – one of the three major credit bureaus in the US – has rolled out its own patent-pending AI model, called NeuroDecision.
In all likelihood, it’s only a matter of time before all credit decisions are made, or at least aided, by AI. Changes are inevitable. But given MAS’ track record in managing Singapore’s financial system, we believe it that AI adoption here will be as fair as possible. Singaporeans should have little to be concerned about – especially when it comes to home financing.
But with or without AI, PropertyGuru will be here to help Singaporeans with their home financing needs. If you’re looking for personalised advice on anything home financing, just fill out this form and one of our professional Home Finance Advisors will get in touch. And for more detailed guides and articles on all things home financing, check out our Home Financing Guides.
Disclaimer: Information provided on this website is general in nature and does not constitute financial advice.
PropertyGuru will endeavour to update the website as needed. However, information can change without notice and we do not guarantee the accuracy of information on the website, including information provided by third parties, at any particular time.Whilst every effort has been made to ensure that the information provided is accurate, individuals must not rely on this information to make a financial or investment decision. Before making any decision, we recommend you consult a financial planner or your bank to take into account your particular financial situation and individual needs.PropertyGuru does not give any warranty as to the accuracy, reliability or completeness of information which is contained in this website. Except insofar as any liability under statute cannot be excluded, PropertyGuru, its employees do not accept any liability for any error or omission on this web site or for any resulting loss or damage suffered by the recipient or any other person.
This article was written by Ian Lee, an ex-banker turned financial writer who hopes to use his financial background and writing skills to help raise people’s financial literacy levels – a necessity in our modern world”
