Complexity tenet | Implications | Recommendations for future research |
Complexity approach. | Generation, identification and explanation of new types of knowledge that holds the world as inherently complex (rather than simple or complicated). | Recognise the world as inherently complex. More relevant research questions making use of qualitative, quantitative and mixed-methods approaches. |
Provided in open systems. | Interventions are influenced by context which is fluid and in flux, as well as potentially many other non-intervention factors. | The limitations of randomised controlled trials and ecological randomised controlled trials should be acknowledged, and more relevant research methods considered. |
Have stratification. | A complex interplay between the individual and their behaviour, as well as the physical and social environment. | Interventions must be described comprehensively, and the interactive, generative effects of components better understood. |
Demonstrate fluidity. | Stratified open systems are always becoming and are thus in flux. | Understand the interconnection and impermanence across stratification and within open systems. Accept fluctuation as a norm and embrace inherent diversity (heterogeneity) as a key feature of complexity. |
Have non-linearity. | Interventions affect outcomes indirectly. | The multifaceted, fluid and flux nature of interventions and their contexts must be accounted for, researched and better understood. |
Have emergent properties. | Interventions can create powers not inherent in the intervention itself. The whole is more than the sum of its parts, and irreducible to these parts. | The manner in which interventions generate powers which affect outcomes needs to be recognised and elicited. |
Have feedback loops. | Interventions affect themselves and reorganise future actions. | Interventions must be described comprehensively and the interactive powers and effects researched and understood, particularly over time and across space. |
Demonstrate improbability. | Intervention outcomes are uncertain, and in some cases unintended, unpredictable and unknown. However, even if outcomes are uncertain, they are not likely to be entirely random. | Instead of controlling for improbability, a complexity lens provides contingencies for facilitating better understandings through studying demi-regularities, and the ability to evolve, learn and adapt. |
Produce demi-regularities. | Intervention outcomes should be understood as somewhat patterned. | Relinquish focus on the false dichotomy of whether an intervention ‘works’ or ‘doesn’t work’. A better, more relevant question is: What works, for whom, when, why and how? |
Evolve, learn and adapt. | Interventions work differently and have different effects over time. | Multiple follow-up evaluations are needed to understand the various ways the intervention affects outcomes. |