Seite - 17 - in Applied Interdisciplinary Theory in Health Informatics - Knowledge Base for Practitioners

characteristics) are emergent properties of the system-as-a-whole. The complexity of modern systems is such that solving one performance issue or bug may introduce others and a holistic perspective on the system together with a deep respect for the complexity of its internal structure becomes essential. As our computer systems have become more complex, they have become, following Boulding’s Level 7, more human. Holistically they can display emergent properties of being buggy, annoying, slow stubborn, inflexible – to the extent that we may find ourselves shouting “stupid computer” at them or complaining about them as though they were a troublesome colleague. From a GST perspective, none of this should be a surprise - most health informatics systems fit comfortably into Boulding’s definitions of Level 5 and above and may have many of the characteristics of Level 7, and perhaps Level 8 too. Especially as modern advances in computing such as AI, neural networks, distributed systems and edge computing increasingly follow biologic models of systems of competing sub-components. The result is that even their designers cannot know exactly how they work. For healthcare this presents an unusual problem: should clinicians trust a computer system that no-one can adequately explain? Medical devices have been regulated on the basis that their programming is rules-based (GST Level 3 and 4) but complexity in general and medical AI in particular have advanced computing well beyond these levels. GST may be needed to help regulation, legislation, the professions and society adjust to human-level computer-based systems. One significant difference between all computer systems and all biologic systems is the relationship with data. Biologic systems process and act on information and store useful information and successful responses to it as knowledge for future reference, and they have used this learning system process to evolve successful survival skills over many thousands of years. Our current computing systems are an awkward fit with GST; they are less than 70 years old and they work differently. Specifically, they can and do store huge amounts of raw data and it is their reliance on data, rather than information and knowledge that can make them appear “stupid”. Future, bio-inspired computing may evolve similar intelligence but for now the key opportunity for organizations is to mine the wealth of big data stored within legacy computer systems. In healthcare, data mining of electronic health records is seen as having the potential to transform our understanding of medicine [8]. Locked away in these records is the history of millions of clinical encounters and their successful or unsuccessful outcomes. 2.3. Applications of GST in Learning Health Systems In health informatics, there has been growing interest in Learning Health Systems, a phrase coined by Charles Friedman [9] in the USA which envisaged rapid learning based on a federated, national approach to exploiting EHR data gathered by different US healthcare providers. More generally, Learning Health Systems are seen as organization- wide or pan-organizational regional and national systems that deliver healthcare to a large population. In Friedman’s vision there is a symbiotic relationship between the health provider system and the health information systems that it uses. The Learning Organization concept was developed from systems theory by management theorists, notably Peter Senge [7]. In learning organizations, systems approaches that reward effective learning are embedded within management culture at all levels of hierarchy. The organization is seen as organic with structures evolving through continuous learning to meet changing environments and ensure survival in a fast paced, ever changing world. O.Johnson /GeneralSystemTheoryand theUseofProcessMining to ImproveCarePathways 17