The partnership that initiated and implemented the pilot is between the University of Limpopo’s Community Engagement Unit and the Waterberg Welfare Society (WWS 2017). The partnership has been reported on in detail elsewhere and so only a brief overview is provided next (Burman & Aphane 2016a; 2016b).

In the last quarter of 2013, discussions between the community–university partners identified one challenge that appeared to be ‘wicked’. The challenge was the legacy of the Abstain, Be faithful, Condomise campaign (ABC) which – despite being phased out by the National Department of Health (NDoH) – continued to frustrate the work of WWS (Burman & Aphane 2016a). Because of the continued frustration that WWS was reporting about the influence or influences of the ABC-legacy, reducing its impact became the practical focus of the partnership. The partnership agreed that the challenge was a ‘wicked problem’ and should therefore be tackled by attempting to build resilience to the challenge, rather than trying to solve it.

Vision 90:90:90 means that the global community:

Vision 90:90:90 has been made possible by the unprecedented medical breakthroughs of the previous decade. These breakthroughs transformed HIV from being a life-threatening condition to a chronic disease if people living with HIV strictly adhere to antiretroviral therapy (ART) (Teeraananchai et al. 2016).

However, despite the unique opportunities afforded by the biomedical breakthroughs, UNAIDS argued that (1) ‘it is necessary to invest in communities … [to] ensure that the AIDS response is people-centred and take it to new levels’ and (2) ‘we have to come up with a biosocial response and not just a medical response to end the AIDS epidemic’ (UNAIDS 2014b:300, 305). They also argued that the new biosocial response had to be capable of incorporating the ‘complex life challenges’ experienced by many people at risk of HIV and AIDS (UNAIDS 2014b:118) which required that there should be a focus ‘not just at the national level but also at the local level’ (UNAIDS 2014b:301).

In 2010 a supplementary issue of AIDS Care included a summary of four dominant biosocial HIV management strategies (Campbell & Cornish 2010), as summarised in Table 1.

Campbell and Cornish (2010:1570) argue that one of the reasons that the ‘first and second-generation’ HIV-related interventions failed to achieve the intended impact is because they were often ‘conceived by external experts’ and imposed on communities in ‘top-down’ ways. As a result, they:

This argument is based upon lessons learnt from global experiences and gave rise to the ‘third and fourth-generation’ approaches that were influenced by the argument that ‘community involvement is a vital pre-condition for creating “health-enabling” social environments’ (Campbell & Cornish 2010:1570). However, a literature review of the impact of community mobilisation in middle- and low-income countries on HIV prevention indicated that ‘no significant results were found for reductions of HIV incidence or prevalence’, although there was ‘suggestive evidence for sex worker groups’ (Cornish et al. 2014:2128). These findings were mirrored by the South African National HIV Prevalence, Incidence and Behaviour Survey, 2012, which presented evidence of the poor track record in relation to HIV incidence rates in South Africa during the previous decade (Shisana et al. 2014).

More recent data also indicate that the biosocial progress towards ending AIDS are not being achieved quickly enough (Mayer, Shisana & Beyrer 2016) – with the exception being new infections in children because of the gains made in prevention of mother-to-child transmission (UNAIDS 2016b). Despite the opportunities that the biomedical innovations are contributing to the ambition of ending AIDS (Vella 2015), UNAIDS continues to emphasise the necessity of reinvigorating biosocial HIV prevention strategies to reinforce the biomedical opportunities (UNAIDS 2016a). This is reinforced by other commentators who argue that the biosocial strategies ‘must be regarded not as complementary, but as integral, to biomedical interventions’ (emphasis added, Govender & Poku 2016:iv).

From the pilot perspective, these findings and arguments prompted a question about whether the types of biosocial innovations that could contribute to ending AIDS should be generated from within the dominant biosocial HIV management paradigm, or whether a disruptive, paradigm shifting innovation was justified. In order to gauge whether, or not, a disruptive innovation was appropriate, a perspective based on broader developmental discourses was developed and applied during the pilot.

Cole (1999:222) argues that ‘what we believe the world to be, ontology, which defines the questions we wish to ask, determines how we study and understand existence, epistemology’ (emphasis in original). For Cole, there are three dominant perspectives that influence the ontological, hence epistemological, parameters of developmental enquiry. The three perspectives include ‘independent actors’, whose potentials are biologically endowed; ‘dependent actors’, whose potentials are a product of ‘adaptive socialisation’ and ‘interdependent actors, whose potentials evolve as emergent properties with social experience’ (emphasis added, Cole 2003:327–332).

These three categories were used as a heuristic to gain insights into the variable perspectives about beliefs in human potentials – and possible biases – within the first- to fourth-generation approaches to HIV management. The heuristic suggests that the first- and second-generation approaches to HIV management are biased towards the independent actor category because they relied on the assumption that people would respond to new knowledge, independent of societal influences. The third- and fourth-generation approaches to HIV management appeared to be biased towards the dependent actor perspective because social environments were considered to either enable or constrain ‘effective community mobilisation’ (Campbell & Cornish 2010:1574).

This brief summary of the assumed potentials of people in communities to engage with the HIV and AIDS epidemic generated two areas that required further consideration. Firstly, if it had been assumed by the designers of the first- to fourth-generation approaches that the potentials of people living in communities are restricted to either the independent or dependent actor category, was this a discrete bias that was adversely influencing outcomes? Secondly, if, as Cole suggests, people influence the social environment – and the social environment influences people and their behaviours – how could this abstract, interdependent concept be built on to improve outcomes during the pilot? The possibility of moving beyond the ontological parameters of the ‘first to fourth-generation’ approaches to HIV management (Cole’s independent and dependent categories) provided one meta-theoretical focus of enquiry for the pilot to determine if additional impact could be derived from applying the interdependent ontological perspective.

The other meta-theoretical query was linked to the association between wicked problems and non-linear complexity, hinted at by Strogatz (2003:182) who argued that ‘every major unresolved problem in science – from consciousness to cancer to the collective craziness of the economy, is non-linear’. In the context of HIV and AIDS, this comment is underscored by Brown et al. (2015:1) who argued that if we rely exclusively ‘on the current linear but disconnected approaches to [HIV] intervention research and evidence we will miss the potential to achieve and understand [complex] system-level synergies’.

In 2008, the incumbent Executive Director of UNAIDS argued that ‘HIV prevention must be able to deal with complexity …. [and] tools need to be designed to capture these dynamics’ (Piot et al. 2008:853). Despite this argument, the literature suggested that there had been very few attempts to address biosocial HIV management using mechanisms associated with complexity theory (for one rare example, see Auerbach, Parkhurst & Caceres 2011). The absence of literature relating to the non-linear complexities that influence biosocial HIV management strategies seemed to be an anomaly in the face of a statement in the Gap Report that in ‘many instances, marginalised populations face complex life challenges, risks and obstacles on multiple fronts’ (emphasis added, UNAIDS 2014b:118) and committed 12 chapters to explain the complex challenges experienced by those populations (UNAIDS 2014b:119–279). This reinforced the idea that attempting to explore the ‘wicked problem’ identified by the community partner using a complexity perspective had potential merit.

The earliest writing about complexity is generally associated with a short essay authored by Warren Weaver (1948). In the period imbetween, complexity has developed into a fragmented international field of scholarship. For the sake of brevity, only the aspects of complexity that have influenced the development of the Framework are outlined in Table 2.

Table 2 highlights some of the dominant characteristics of complex systems. One of the most striking characteristics of complex systems is that they do not produce entirely random, nor entirely predictable, outputs or outcomes. This means that whilst it is not possible to produce replicable results that fit the criteria associated with evidence informed best practices, it is possible to manage them (Ramalingham 2013; Snowden & Boone 2007). The other aspects of complex systems that are relevant include (1) complex systems are made up of multiple agents that interact through feedback loops (Dunn et al. 2016); (2) agents are understood to mean any entity that influences the system (Sturmberg & Martin 2009); (3) the outcomes, or outputs, associated with complex systems emerge from the feedback interactions of the agents within the system (Shiell, Hawe & Gold 2008), and (4) the feedback processes that maintain the system are self-regulating (Pincus & Metten 2010), self-organising (Lee 1997) and self-sustaining, or – to use jargon from complexity, autopoietic – (Chettiparamb 2007). Emergence from a complex system is thus mediated by discrete and indeterminate interactions of the agents within the system which is represented in Figure 2 using an iceberg metaphor.

The emergence is a result of the feedback interactions between both the surface descriptor – that is the characteristics of a phenomenon, such as adherence to medication – and the submerged properties that co-constitute the feedback dynamics between the agents within the system that generate and sustain the characteristics of the phenomenon. Consequently, management of complex systems requires altering the submerged feedback dynamics between the agents in order to change the properties of the system, which in turn, alters the subsequent emergence (Coleman et al. 2007).

The application of complexity theory to the pilot enabled the HIV and AIDS epidemic to be formulated as the consequences (outputs and outcomes) that emerge from the feedback dynamics within, and between the submerged components of the ‘human-virus-environment’ that sustain the HIV and AIDS system (Burman, Aphane & Delobelle 2015:15). This formulation began a process of considering that one of the likely mediators of the outcomes of the pilot would rely upon initially working with the discrete, submerged, properties of the dynamical system, rather than focusing exclusively on the surface descriptors.

To achieve this required insights into how the submerged feedback dynamics influence emergent outputs and outcomes. The dynamical feedback processes within complex systems are influenced by strange attractors (henceforth ‘attractors’), which means that attractors represent one focus of the management strategy (Snowden & Boone 2007). Attractors represent recurring convergences of agents towards multiple equilibria zones within a complex system which influence emergence (Pincus & Metten 2010; Sturmberg & Martin 2010). ‘Strange attractors are reflected in patterns of behaviour, that is, shapes in space or movements over time, which are never exactly repeated but are always similar to each other’ because they are fractal in structure (Stacey 2003:44). Consequently, as attractors are discrete forces that mediate self-similar emergence, they represent a loci of change opportunities when managing complex systems – and thus became a strategic analytical focus during the pilot (Burman & Aphane 2016a).

From the pilot perspective, it was not only necessary to gain insights about how attractors – dominant potential loci of change – influence systems, it was also necessary to identify a mechanism to simultaneously manage both non-linear and linear emergence during the pilot. The primary source that influenced the development of a mechanism to manage emergence was a theoretical article in the Harvard Business Review called A leader’s framework for decision making (Snowden & Boone 2007). The authors of the article introduced a management heuristic called the Cynefin framework that enables leaders to identify both non-linear and linear challenges and provides a management response typology for each category (Figure 3).

The two domains on the right-hand side of the Cynefin framework – complicated and obvious – represent ordered, predictable, linear decision-making contexts and the two on the left-hand side – complex and chaos – represent unordered, unpredictable, non-linear decision-making domains. The Cynefin framework has been applied in multiple contexts (Dettmer 2011; Elford 2012; Sturmberg & Martin 2008; Van Beurden et al. 2013) and the intention was to co-opt it into the HIV challenge that the partnership was focusing on.

Responding to the different domains required that the partnership understood how to implement each management response typology. The best, good and novel practices within the obvious, complicated and chaos domains were easy to understand. However, the emergent practice in the complex domain required further investigation. Snowden and Boone (2007:72) argue that in complex scenarios, it is only possible to ‘understand why things happen only in retrospect’. However, they also argue that when attempting to identify opportunities to manage complex situations, it is possible to identify ‘instructive patterns’ if ‘safe to fail’ experiments are undertaken (Snowden & Boone 2007:72).

Safe to fail experiments – contra fail–safe strategies – represent low risk, ethically sound experiments of which about 50% are expected to fail, but the learning from the failure can be exploited to improve the more successful experiments (Dickens 2012; Huertas 2014). When working with complexity, this is a viable strategy for developing future management opportunities because complex systems are robust and have the capacity ‘to reorganise and recover from change and disturbance without changing to other states’ (Ahern 2011:341). The notion of a safe to fail management response to complex challenges was therefore integrated into the Framework design.

Non-linearity – inputs which produce disproportional outputs – is a useful indicator of a complex system. In the case of HIV and AIDS management, ‘billions of dollars have been poured into HIV and AIDS interventions’ (Campbell & Cornish 2010:1570), yet incidence rates in South Africa remain stubbornly high (Shisana et al. 2014). The disproportionately small impact from the billions of dollars suggests that the HIV and AIDS epidemic is systemically robust in the face of attempts reduce risk and is an indicator of non-linear complexity. Once it has been agreed that the system contains some non-linear complexity it is necessary for the partnership to agree an end-condition.

An ‘end-condition is imagined and not a specific end-state’ because complex systems are unpredictable; thus, it is not possible to determine the precise ‘end-state’ – but it is possible to predict the desired parameters of emergence. In this instance, because the 90:90:90 ambition represents reducing the aggregate global viral load to the lowest possible level, it was decided by the partnership that the end-condition for the pilot would be a ‘reduced aggregate community viral load’ (Burman & Aphane 2016a:95).

The next phase involves destabilising the environment within which the wicked problem is situated and has been labelled the primary probe (Figure 4). It is possible to ethically destabilise the system because complex systems are inherently unstable, yet robust enough to recover, through adaptive processes that may ‘create new order’ (Dunn et al. 2016:4). In this instance, the primary probe was an educational package called ‘A-3B-4C-T’ that focused on the human–virus–environment interactions (Leadership Training Resource Centre 2014). The A-3B-4C-T educational package was delivered to WWS’ leaders to provision them with new perspectives about the HIV and AIDS epidemic (Burman & Aphane 2016a).

If the primary probe has been effective, a period of systemic re-organisation occurs as people affected by the destabilisation begin to make sense of the changes and act on them. This process reflects a series of indeterminate iterations that can be associated at a theoretical level with frame theory (Goffman 1974), including ‘unlearning’ existing frames (Rogers et al. 2013:4) and reconstructing new frames which are influenced by the primary probe. This process involves sense-making (Dervin 1998) and closely resembles theories associated with experiential, adult learning (Wals 2007). In this instance, the systemic re-orientation catalysed new attractors which WWS began to incorporate into their work.

In order to identify the existing and emergent attractors, a qualitative analytical methodology called causal layered analysis – which is designed ‘to deconstruct complex social issues’ (Bishop & Dzidic 2014:13) – was applied 10 months after the A-3B-4C-T educational package had been delivered. In this instance, six attractors that contributed to the end-condition ambition and three that detracted from it were identified (Table 3).

Table 3 provides an overview of the attractors which were influencing the local HIV and AIDS environment 10 months after the A-3B-4C-T educational package was delivered. The emergence of new attractors indicated that WWS was developing emergent perspectives with regard to the ABC-legacy.

The Cynefin framework is a decision-making heuristic that is designed to determine whether or not a challenge is exclusively linear or contains some non-linearity, so that appropriate management decisions can be determined (Snowden & Boone 2007). In this instance, it was also used as a monitoring and evaluation device which was designed to identify movement within the Cynefin framework as an indicator of the utility of the Framework at that stage in the pilot (Burman & Aphane 2016a).

Joint United Nations Programme on HIV/AIDS UNAIDS do not define what the new biosocial response should look like. However, by extrapolating the concepts that contributed to the development of the Framework, the following can be suggested: (1) the introduction of new biomedical opportunities to end AIDS could be generating destabilisation within the South African HIV and AIDS environment; (2) the likelihood is that the complex HIV and AIDS environment is consequently undergoing a period of systemic re-organisation in response to the biomedical opportunities; and (3) the likelihood is that new attractors are emerging at multiple scales.

If it was possible to identify and respond to the emergence, then the South African NDoH and partners could respond with increased precision. However, this would require a dual focus on targeting not just people at risk (the surface descriptor), but also the discrete, attractors that mediate – or influence – the patterns of risk (the properties of the complex system). This is explored in more detail below.

The South African HIV and AIDS environment is undergoing a profound shift – a form of destabilisation at a national scale – which is being enabled by the NDoH’s commitment to Vision 90-90-90. This destabilisation is primarily being driven by the increased coverage of biomedical technologies and is likely to produce a broad portfolio of systemic, biosocial ripple effects along the entire treatment and care continuum (Whiteside, Cohen & Strauss 2015). Some of these ripple effects will generate desirable emergent outcomes and contribute to achieving UNAIDS’ end-condition of ending AIDS by 2030. However, there will also, inevitably, be undesirable emergent outcomes that will undermine the ambition. One possible example includes the way in which the biosocial ripple effects associated with pre-exposure prophylaxis (PrEP) medications could generate dual ripple effects: ‘PrEP has the capacity simultaneously to increase judgement and stigma about sexual behaviour and to decrease fear and stigma in sexual encounters’ (Brown et al. 2015:4).

Following the logic of the Framework – and assuming that there is consensus that the HIV and AIDS environment contains some non-linear, biosocial complexity – targeting the emergent, systemic re-organisation of the national HIV and AIDS environment as it responds to the destabilisation that is being generated by biomedical opportunities is a pragmatic management response. This is briefly discussed subsequently.

Ideologically, UNAIDS – and by default the South African NDoH – takes an evidence informed position that it is necessary to target specific, at risk, populations. There is a pragmatic logic to this approach if the challenge is exclusively linear. However, as the epidemic is now becoming more ‘complex’ and ‘nuanced’ (UNAIDS 2015:178) and the 90:90:90 window of opportunity is closing, there is also a logic to respond to the non-linear factors that influence the South African HIV and AIDS environment. This would require identifying emergent attractors that affect the trajectory of the epidemic as affected people respond to the ripple effects of the biomedical opportunities.

Once the attractors that contribute to ending AIDS have been identified they can be reinforced and the attractors that do not contribute to ending AIDS can be dampened so that the submerged feedback dynamics – the properties of the system – which influence the global end-condition of ending AIDS by 2030 are reinforced appropriately. By doing this, it will be possible to consolidate the social practices that contribute towards achieving the 90:90:90 end-condition of ending AIDS as they are developing – and simultaneously dampen social practices that detract from the 90:90:90 end-condition. This methodological design would be inclusive of both the surface descriptor and the submerged properties of the HIV and AIDS environment which suggests that the response would be (1) likely to make sense at localised scales; and (2) be sustainable, because the focus of the intervention would be the whole system – rather than just the surface descriptors.

The implications of this is that the new biosocial response to the changing HIV and AIDS environment needs to be responsive to both serendipity and rational opportunities to target both linear and non-linear emergence. By targeting the nuanced, emergent shifts so that they are being continually directed towards achieving the end of AIDS could represent an innovative biosocial addition to the 90:90:90 armamentarium. Expanding the application of the Framework – especially if it was augmented by pattern detection software such as SenseMaker^® – could represent an opportunity to begin this process. From this perspective, the Framework does represent a nascent biosocial response that can make localised sense in variable contexts.