This white paper has been developed by the Task Force to explore the ethical considerations of AI.
We have grouped the ethical considerations into two themes:
• Data Input and Design (e.g. AI acting on data set bias), and
• Understanding and Control (e.g. a lack of transparency of AI related decisions).
We have made recommendations for how firms can address the ethical considerations of these two themes. In summary firms should:
• Apply a critical view of the data sets used for each AI application and perform specific quality control checks from design through to operation;
• Train design teams and individuals to be aware of, and mitigate, biases within the function and design of individual AI applications;
• Assess what level of ‘explainability’ is necessary for each AI application and take this into account in its design; and
• Ensure that all AI applications are subject to a suitable control framework and audit process throughout their lifecycle.
AI is a fast-developing technology which is already being applied to specific uses within firms, as well as moving towards larger-scale deployments. Continuing discussions and information sharing on the ethical considerations of AI will be critical for the development of the capability within capital markets.