Technology

Could Artificial Intelligence Transform Credit Scoring?

Could Artificial Intelligence Transform Credit Scoring?
Table of Contents
  1. The Flaws of Traditional Credit Scores
  2. Artificial Intelligence in Lending
  3. Concerns Over Bias and Regulation
  4. Alternative Approaches and Industry Shifts

Securing a mortgage, car loan, or mobile phone contract often depends heavily on an individual’s credit score. However, experts and consumers point out that traditional credit scoring systems can be outdated, prone to errors, and disadvantageous to certain groups, prompting a look at whether artificial intelligence can offer a better way to assess creditworthiness.

The Flaws of Traditional Credit Scores

Nail technician and content creator Sophie Hope experienced the severe impact of traditional credit reports firsthand when a lender abruptly denied her mortgage application. She discovered she was a victim of identity fraud after someone opened catalogue accounts in her name, transferring debt to a secondary finance company that accumulated over three to four years without her knowledge and ruined her credit score.

Funmi Olufunwa, a consumer finance lawyer and founder of financial education business Hoops Finance, noted that people are overly obsessed with credit scores in the UK, where three major firms—Experian, Equifax, and TransUnion—produce scores using similar data. Olufunwa explained that a credit score is backward-looking and focuses on past events rather than accurately predicting the future or assessing actual affordability.

Traditional systems also disadvantage young people and recent immigrants who lack a built-in credit history, often referred to as thin-file applicants.

Artificial Intelligence in Lending

With vast amounts of financial data available, some lenders are turning to artificial intelligence to determine who is safe to lend to. In the United States, lending decisions frequently rely on FICO scores, but AI-powered lending firms like Upstart and Zest AI provide alternative options for evaluating creditworthiness.

Upstart stated in a release that machine-learning models analyzing more than 2,500 variables allow them to issue significantly more loans at lower prices to borrowers across all demographic groups. Sarah Tucker, a mortgage expert and founder of The Mortgage Mum, stated that AI has the potential to help lenders analyze broader information sets, spot patterns quickly, and build individual affordability and risk profiles. Tucker believes this offers a genuine opportunity to improve financial inclusion for thin-file applicants instead of relying on rigid rules.

Concerns Over Bias and Regulation

Despite the potential benefits, the integration of AI in everyday financial decisions brings significant concerns. Tucker emphasized that consumers must understand what data is utilized to make decisions about them and retain the ability to challenge incorrect outcomes. Sandra Wachter, a professor of Technology and Regulation at the University of Oxford, pointed out that credit scores have remained contentious for decades and argued that AI does not account for existing social inequalities, risking the replication of past biases.

Wachter stressed that real transformation requires accountability alongside automation, urging firms to enforce existing regulations and build new guardrails. Without proper safeguards, discrimination will be fortified and enhanced, according to Wachter. In response, Upstart stated that it has invested in automated evaluations to identify bias and integrated them into its model training process to ensure the system becomes more accurate and fair over time.

Alternative Approaches and Industry Shifts

Alternative models have also emerged outside of traditional credit agencies. Evangeline Atkinson faced rejection for a phone contract while at university due to being completely invisible to the major credit reference agencies. In response, Atkinson co-founded a credit reference agency called Noggin HQ, which was authorized by the Financial Conduct Authority (FCA) in June 2021 to help consumers access affordable credit by analyzing current account transactional data without using AI.

Atkinson warned that using AI carries the risk of inbuilt bias because sensitive data could easily lead models to infer protected characteristics. Meanwhile, traditional credit scoring agencies are beginning to experiment with AI themselves. Experian announced a partnership with OpenAI to let users check credit scores inside ChatGPT, Equifax launched its Equifax Amplify AI platform for sharper credit assessment and fraud detection, and TransUnion integrated Google’s Gemini models into its analytic data assets.

Olufunwa noted that companies are finally realizing traditional credit scoring is inherently backward-looking and that better predictors of the future exist. Whether older and younger generations will feel equally comfortable with AI determining their future credit ratings remains to be seen, as tech-savvy younger consumers are generally more accustomed to their data being used in this manner.

You May Also Like

Leave a Comment

Your email address will not be published. Required fields are marked *