The increasing adoption of artificial intelligence (AI) in the financial sector has brought about unprecedented benefits, including enhanced operational efficiency and improved decision-making. However, as AI systems become more autonomous, the need for human oversight has become more critical than ever, according to Nolwazi Hlophe, senior specialist: fintech at the Financial Sector Conduct Authority. Hlophe was speaking at the recent ITWeb AI Summit, where she emphasised the importance of human integration in AI systems, particularly in finance.
While AI has the potential to revolutionise the financial industry, over-automating AI systems can lead to inefficiencies and risks. Without human oversight, AI systems may raise false alarms, miss critical transactions or perpetuate biases and discrimination. In finance, where trust and accountability are paramount, human intervention is essential to ensure that AI systems operate ethically and responsibly, noted Hlophe.
She highlighted the importance of the "human-in-the-loop” (HITL) approach, which integrates human feedback at various stages of AI and machine learning development processes. “This approach enables human experts to correct errors, provide labelled data and validate AI outputs, thereby enhancing the accuracy, reliability and adaptability of AI systems.”
Applications of human-in-the-loop in finance
Hlophe said the benefits of HITL approaches are evident in various financial applications, including fraud detection, credit scoring and chatbot customer service. In fraud detection, AI might raise false alarms as opposed to picking up real fraudulent transactions, and this is where a human being will come in to monitor that exchange, noted Hlophe.
“Credit applications and scoring are particularly vulnerable to bias and discrimination, making human oversight crucial for maintaining trust in the financial system. Additionally, human-in-the-loop approaches have proven valuable in chatbot-based customer service, where complex queries require human intervention.”
To ensure effective human oversight in AI systems, Hlophe recommended the following best practices, defining clear roles and responsibilities by ensuring that both AI systems and human operators understand these:
- Utilise advanced tools and technologies: Implement tools that facilitate seamless interaction between humans and AI, such as user-friendly dashboards and real-time monitoring systems. These technologies enhance the efficiency of human oversight.
- Invest in training and development: Provide comprehensive training programmes for financial professionals to enhance their understanding of AI systems and interpret AI outputs effectively.
- Establish continuous feedback mechanisms: Create processes for ongoing feedback from human operators to improve AI models.
“We need to make sure that we work together with this technology to improve its function and its outcomes.”
Share