Advocating systems approaches outside of engineering: reporting on research with the UK insurance sector towards the adoption of trustworthy AI-enabled systems
conference contributionposted on 11.12.2020, 11:29 by Paul TimmsPaul Timms, Melanie KingMelanie King
Part of INCOSE’s SE vision 2025 is to expand the application of Systems Engineering approaches across industries. One opportunity is the development and deployment of AI-enabled services by enterprises worldwide, as societal expectations for acceptable AI use present significant challenges for policy makers, regulators and system designers. In this paper, we report on an ongoing research project with the UK insurance sector, investigating the application of systems approaches in the adoption of trustworthy AI-enabled systems. We aim to showcase the relevance of a systems approach when addressing impending consumer protections for AI use, referenced as principles for ethical, responsible, or trustworthy use of AI and data technologies, amongst others, dependent on publication and perspective adopted. We explain how the UK insurance industry, and underpinning platform technologies, can be considered a complex system of systems, and how Systems Thinking worldviews can bring coherence. This research builds upon two substantial bodies of work that have synthesised policy and advisory literature (combining governmental, inter-governmental, private sector and civil society stakeholders) on the expectations for acceptable AI and data use across the world. We describe how Systems Engineering practices can be applied to integrate trustworthy design principles into a complex socio-technical environment, and discuss why capability engineering, at the insurance sector level, can help insurance enterprises and industry regulators address ethical / responsible / trustworthy-by-design criteria. We comment on the challenges and limitations we have encountered during this process and discuss the next steps of this research.
- Mechanical, Electrical and Manufacturing Engineering