Artificial intelligence (AI) enables innovations in digital finance that can boost efficiency, cut costs, and expand consumer reach. AI provides scalable ways to determine the creditworthiness of individuals and businesses that historically lacked identification, collateral, and credit history. And by automating processes, AI can enable higher volumes of low-value transactions that make harder-to-reach segments more viable customers. However, while AI presents tremendous opportunities, deploying AI also comes with significant risks – one being inequitable outcomes for marginalized consumers, especially women. When applied at scale, the harms caused by AI counter the financial inclusion goals of many impact investors and their portfolio companies. While impact investors are well positioned to take a proactive role in supporting companies in building and deploying equitable AI systems in inclusive finance services, they often do not have the tools required to assess and evaluate the fairness of algorithms. Recognizing this gap, the Center for Financial Inclusion (CFI) developed a practical guide for impact investors to help them identify harmful AI gender bias.
The Investing in Equitable AI for Inclusive Finance: A Risk Management Guide for Impact Investors gives an overview of the use cases of AI in inclusive finance and some of the drivers of harmful AI bias towards women. It also presents a snapshot of the state of practice in bias identification and mitigation and then provides a user-friendly checklist with an actionable set of questions to help impact investors understand the use of AI among their investee companies.
The accompanying brief, Prompts for Equitable Artificial Intelligence in Inclusive Finance: Strengthening the Industry Conversation, presents some of the key challenges with building accountability and transparency for AI in inclusive finance and discusses how CFI’s new guide can help raise awareness around the risks and costs of gender bias in AI. The brief also offers opportunities to strengthen the industry conversation on equitable AI, including advancing research, supporting consumer advocacy and journalism, and operationalizing data rights for consumers.
This work is the result of USAID’s Equitable AI Challenge. Implemented by DAI’s Digital Frontiers, the challenge aimed to identify innovative approaches to address artificial intelligence inequitable outcomes. The Equitable AI Challenge asked for proposals that critically consider holistic and creative approaches to identify and address gender biases in AI systems within global development contexts. This work is the result of CFI’s winning proposal.