"Loan Screening with Artificial Intelligence? Increasing Transparency of the Black Box"

Joint Forum of Bank of Korea and Statistical Society on the 15th
Enhancing Reliability through Advancement of 'Explainable AI'

It has been suggested that to integrate AI (artificial intelligence) into financial sectors such as loan screening, explainable AI should be utilized to enhance the transparency of the black box.

At the Korea Statistical Society-Bank of Korea joint forum held on the morning of the 15th at the Bank of Korea headquarters in Jung-gu, Seoul, under the theme "Expanding and Applying Economic Statistics in the Era of Explainable AI," Professor Kim Seonghwan of Konkuk University's Department of Statistics stated, "When AI is introduced to determine the eligibility for credit loans, if a customer's loan is rejected and the customer requests the reason for the rejection, the bank can only respond that 'the latest AI algorithm made the decision based on big data,' raising concerns about the reliability of the black box."

Traditional AI has the limitation of not being able to provide clear evidence because it mechanically learns from big data to produce results. The concept developed to overcome this drawback is 'explainable AI.' Professor Kim explained, "Explainable AI is a system that provides understandable explanations of the AI structure and the validity of decision-making," adding, "Its purpose is to reduce the gap between AI algorithms and human trust."

Professor Kim emphasized that to practically utilize AI in financial and economic statistical forecasting, discussions beyond academia and into the field are necessary. He highlighted three challenges: "Financial data is difficult to collect, it contains socio-psychological elements that are hard to explain with only quantified indicators, and the social ripple effects of using AI algorithms are significant."

AI utilization in the financial sector is still in its early stages, and social trust regarding the transparency and fairness of the technology is insufficient. Dr. Cho Hyunseon of LG CNS also stated, "Although AI is being introduced in various fields, due to institutional shortcomings, lack of reliability, and insufficient quality data, it is not being used as much in the financial sector as expected, and development has been limited accordingly," adding, "It is important to make model decisions transparent to provide reasons that external stakeholders can accept."

Dr. Cho said, "When interpreting results produced by explainable AI, subjectivity can intervene, leading different viewers to draw different conclusions," and added, "It should be used as an auxiliary indicator for decision-making, and applying ensemble methods (techniques that combine results from multiple models to produce a final outcome) with various AI models is also desirable."

On the 15th, participants are taking a group photo at the 2023 Korea Statistical Society-Korea Bank Joint Forum held at the Bank of Korea. Photo by Bank of Korea

This content was produced with the assistance of AI translation services.