The reemergence of systems research in dietary health

Though relatively new to public health and food, systems thinking approaches are rooted in a deep history, perhaps many centuries old. In more recent history, systems approaches were commonly applied to cosmology^{(Reference Von Humboldt1)} and engineering; however, arguments for the application of General System Theory were introduced in the 1950s by biologist Ludwig von Bertalanffy^{(Reference Von Bertalanffy2)}. Not long after, Donella Meadows pioneered systems thinking approaches for global environmental challenges^{(Reference Meadows3)}, and creators of second-order cybernetics (another branch of systems thinking) emerged, such as cultural anthropologist Margaret Mead^{(Reference Mead4)}. Using systems approaches has since become commonplace in health, public health and food systems and is advocated by government agencies, including Public Health England for obesity reduction⁽⁵⁾. We are learning now that systems approaches can provide tools to help understand issues, constraints and opportunities in order to both consider areas to prioritise and to develop interventions. Importantly, through our understanding of how factors within the system connect and flow, we can develop interventions that do not result in equilibrium (where one policy is diluted by other counteractive actions, (i.e. in a balancing feedback loop)) and avoid or deal with blockages or loopholes that mean our interventions have a lower chance of success. Having this understanding also enables us to consider why an intervention or policy does or does not work. Additionally, systems maps enable us to visualise these relationships, highlighting complexity and demonstrating how altering one variable in a system can influence others.

The obesity Foresight map was one of the earliest examples of systems maps used to highlight complexity within health⁽⁶⁾. This also clearly demonstrated how obesity was caused by multiple interrelating factors across different domains or disciplines. As a result, not only has it influenced the policy landscape, it has forged multi-sector, multi-disciplinary partnerships with experts working across the whole system. Understanding that obesity is complex has moved the narrative away from simplistic views that excess weight gain is a result of poor motivation, to one in which multiple factors interrelate to influence appetite regulation; it can also reduce obesity stigma and shame.

This review paper aims to introduce readers to the application of systems approaches in research within the context of food systems and health, drawing upon an understanding of the traditional/historical origins. We first outline current food system challenges and the urgent need for systems approaches. We then provide examples of how systems research has been used to understand the drivers of food choice and health outcomes, how this understanding can shape intervention or policy development, and methods that can be used to support co-design, evaluation and systems governance.

The broken food system

Food systems are complex, comprising an interplay between macro (higher-level political, economic and social factors), meso (organisations and groups) and micro (individuals) sub-systems. Multiple sectors are impacted along the supply chain, from production, manufacturing and processing, to sales, consumption and waste management^{(Reference Hasnain, Ingram and Zurek7,Reference Parsons, Hawkes and Wells8)} . Each part of this journey can positively or negatively impact on individual and planetary health^{(Reference Larsen, Ryan and Abraham9)}. Effective interventions ought therefore to consider food systems approaches that take into consideration competing priorities of different sectors, to ensure effective and long-lasting impacts^{(Reference Mikkelsen, Novotny and Gittelsohn10)}, even when evaluating stand-alone interventions^{(Reference Sisnowski, Street and Merlin11)}. From a public health nutrition perspective, the global food system currently pushes energy-dense, high fat, sugar and salt (HFSS) and ultra-processed foods onto consumers, leading to a state of overconsumption^{(Reference Barnhill, Palmer and Weston12)}. National Diet and Nutrition Survey (NDNS) data suggests that less than 1 % of the UK population currently adheres to all nine recommendations of the UK Eat Well Guide, whilst only 31 % adhere to five recommendations^{(Reference Scheelbeek, Green and Papier13)}. Further, while the global food system generates sufficient amounts of food to feed the world’s population, we are witnessing unprecedented levels of food insecurity. In the UK, latest figures from the Food Foundation suggest that ∼14 % of households experience food insecurity, rising to 18 % amongst households with children⁽¹⁴⁾. Those facing food insecurity are more likely to experience a double burden of disease, where they are both malnourished and experiencing obesity^{(15,Reference Carvajal-Aldaz, Cucalon and Ordonez16)} .

Issues with our food systems are not equally distributed. Unhealthy diets that are more prevalent due to the availability, affordability and accessibility of energy-rich foods^{(Reference Caspi, Sorensen and Subramanian17,Reference Cummins and Macintyre18)} , affect those on lower incomes disproportionately^{(Reference Cooksey-Stowers, Schwartz and Brownell19,Reference Maguire and Monsivais20)} . As a consequence, there is a 19-year gap in healthy life expectancy in the UK between those living in the most and least affluent neighbourhoods⁽²¹⁾ and at least a two-fold difference in obesity risk⁽²²⁾. The cost of healthy food options exceeds the cost of convenience, ultra-processed foods three-fold^{(Reference Hoenink, Garrott, Jones, Conklin, Monsivais and Adams23)} and healthier options have increased in cost more substantially than unhealthy alternatives^{(Reference Goudie and Hughes24)}. In fact, to meet UK Government dietary guidance, those living in greatest poverty would need to spend 43 % of their disposable income on food in comparison to only 10 % amongst the least deprived⁽²⁵⁾. Additionally, as energy prices have additionally soared, it has become increasingly difficult and costly for individuals to cook healthy and sustainable meals from scratch⁽²⁶⁾.

Fast food outlets are also more prevalent within areas of greater deprivation in comparison to the least deprived⁽²⁷⁾ and evidence highlights how this influences childhood obesity^{(Reference Fraser, Edwards and Cade28–Reference Burgoine, Sarkar and Webster30)}. Retailers have encouraged purchases of unhealthy foods, more so than healthier alternatives^{(Reference Smithson, Kirk and Capelin31)}, through price promotions and the encouragement of ultra-processed food^{(Reference Watt, Beckert and Smith32)}. School food systems are not immune to this, with analysis of the NDNS suggesting that over 70 % of energy intake from lunch eaten at school is from ultra-processed food^{(Reference Parnham, Chang and Rauber33)}.

There is also a clear environmental impact of the food system, which travels alongside the complete food production and supply chain^{(Reference Crippa, Solazzo and Guizzardi34)} and then creates excess waste, both in terms of the food and the packaging it comes in^{(35–Reference Johnstone and Lonnie38)}. In particular, our overconsumption of meat and foods high in fat, sugar or sale (HFSS) and lack of fibre, fruit and vegetables^{(Reference Dimbleby39)}, also contributes substantially to climate change, for example, through greenhouse gas emissions^{(Reference Scheelbeek, Green and Papier13)}. Recent research has suggested that if the UK population adhered to current Government guidance, there would be a significant reduction of approximately 30 % in the overall carbon footprint^{(Reference Scheelbeek, Green and Papier13)}. Reductions in water use by 17 % and reductions in land use by 34 % have additionally been reported⁽⁴⁰⁾. Maintaining a sustainable diet would also decrease risks of ill health and prevent diet-related diseases^{(Reference Scheelbeek, Green and Papier13)}. Thus, the current food system does not encourage the adoption of sustainable, healthy diets.

The urgent need for a food systems approach

The negative impact of our food system on population, economic and planetary health requires urgent attention and regeneration rather than prevention of further harm^{(Reference Doherty, Bryant and Denby41,42)} . Without action, these impacts will worsen as food systems expand to feed almost 10 billion people by 2050^{(Reference Willett, Rockström and Loken43)}. It is now acknowledged that the development of the current global food system, which focused on addressing the singular need to feed a growing population, paid insufficient consideration to how this could impact other elements of the food system and beyond^{(Reference Dimbleby and Lewis44)}. For example, the move to high-yielding monoculture crops supported by artificial pesticides and fertilisers in the later part of the 20th century is credited with preventing starvation but also contributing to disease burden, biodiversity loss and climate change^{(Reference Dimbleby and Lewis44–46)}. Similarly, the move to maximise convenience for economic and social reasons led to a shift in power towards large corporations providing highly processed foods which are also associated with an increase in human disease and global warming^{(Reference Dimbleby and Lewis44,Reference Griffith, Jin and Lechene47–Reference Hall, Ayuketah and Brychta50)} .

Taking a holistic food system approach helps all relevant impacts to be considered, increasing the likelihood of positive outcomes from any interventions^{(Reference Von Braun, Afsana and Fresco51)}. The value of a systems approach has been recognised by the UN and UK government with key strategies being developed in this manner in recent years^{(Reference Dimbleby39,52–54)} . Furthermore, policymakers are starting to leverage the significant levels of public sector food spend to strategically influence the food system to deliver their social, economic, health and environmental mandates^{(Reference Swensson, Hunter and Schneider55–Reference Morgan and Sonnino61)}. This is complex in practice, with economic, environmental, health, social and political agendas varying, and often conflicting, across different stakeholders^{(Reference Parsons62)}. A holistic view is also important as the system may develop its own behavioural pattern which is distinct and separate from that of its parts^{(Reference Meadows63)}. Systems thinking which addresses opportunities and issues holistically, considering the impact of any interventions on the system overall rather than for individual parts in isolation, are also most likely to mitigate undesirable consequences and system dysfunction^{(Reference Parsons62,Reference Meadows63)} . When this is done well, actions on one part of the system can have a beneficial impact on the food system as a whole. For example, changes to public sector procurement policy which have positively influenced commercial behaviours has been evidenced in the UK through the public sector requirement for sustainable fish, and in Denmark through the public sector transition to organic prevalence^{(Reference Dimbleby39,Reference Swensson, Hunter and Schneider55)} .

Taking a visionary approach to food systems

Donella Meadows, an environmental scientist, was a pioneer in using systems thinking to address complex problems from the 1970s. Meadows introduced new audiences to systems approaches and helped people to understand interconnectivity between our actions and ecosystems, whilst inspiring people to use visionary approaches to imagine what a preferred future would look like As Meadows described: ‘In my vision, food is raised and prepared as consciously and lovingly as are children, with profound respect for nature’s contribution as well as that of people’^{(Reference Meadows64)}. Importantly, this contrasted with approaches to improve environmental sustainability that were solely focused on restriction and regulation, and encouraged people to think about what an improved future might look like, as well as doom-laden narratives of dystopian futures.

Visionary approaches are now increasingly applied to help organisations develop strategies, e.g. as part of a Three Horizons futures process^{(Reference Sharpe, Hodgson and Leicester65)}. Again, these have traditionally not been used in public health research. Several recent case studies, however, have applied systems-informed visionary methods to explore what radically different, desired future food systems might look like. For example, the FixOurFood research programme^{(Reference Doherty, Bryant and Denby41)} used multiple Three Horizons workshops and surveys with 114 experts (including representation from 61 different organisations) from across the Yorkshire food system to envision a regenerative future food system for the region, underpinned by a fundamentally different set of values and worldviews to the problematic current food system, and drawing on the Regenerative Lens framework^{(Reference Buckton, Fazey and Sharpe66,Reference Buckton, Fazey and Doherty67)} . The Seeds of Good Anthropocenes project and approach, which supports development of positive future visions based on elements of radical, inspiring initiatives already active in the present, has also been applied in food system contexts around the world^{(Reference Guerrero Lara, Pereira and Ravera68–Reference Sellberg, Norrby and Nowak71)}. The US$2 million Food System Vision Prize, launched by The Rockefeller Foundation in partnership with SecondMuse and OpenIDEO in 2019, rewarded the top bold, actionable visions of regenerative food systems in 2050, recognising the importance of inspiring visions to spur transformative innovation and action in the present⁽⁷²⁾.

Understanding systems

As previously noted, with regards to inherent complexity and interconnectedness of the food system in which multiple components are dependent on each other (and addressing one element in isolation) risks, at best, limited understanding and, at worst, negatively impacting other parts of the system. This is particularly applicable in the case of community and charitable food systems, an evolving and under-researched area and yet increasingly a major source of food for people living on a low income. Community food provision – or community food organisations (CFOs) – are defined as organisations and initiatives that support people in poverty and/or food insecurity to access food for free or at very low cost. This incorporates food banks, soup kitchens, community cafes, social supermarkets, community kitchens and community gardens. Community food provision is an ill-defined and quickly evolving sector, which expanded rapidly during and in the aftermath of the COVID-19 pandemic. Critically, the sector does not exist in isolation but is closely linked to other elements of the food system, such as retailers, and external sectors, such as health, welfare and local government. Understanding how community food organisations connect not only with other community food organisations but also with other elements of the food system and other sectors is critical to understanding the sustainability of CFOs as a whole and illuminating the extent to which they are able to help people in need^{(Reference Graver73)}.

The NIHR PHR-funded ‘Fair Food Futures UK’ project has adopted a systems approach to understand both the different organisational models and components of CFOs and the interconnections between them, and the broader social and economic systems in which CFOs operate^{(Reference Sheard, Previdoli and Burton74)}. Data were collected (surveys, workshops, interviews (local and national), and co-design workshops) to help understand the key organisational components and political pathways for successful/unsuccessful implementation of services provided to achieve inclusivity and accessibility, and move clients away from emergency food aid. The resulting systems map allowed for an exploration of feedback loops and causalities, highlighting potential points of successful intervention, where small changes in one part of the system can directly or indirectly affect other elements, or result in significant wider change. In this way, it enabled an exploration of which decisions, shifts and policy interventions have the potential to change the system in the most effective way for a more targeted use of CFO resources (identifying actions to achieve policy targets, trade-offs and synergies in CFOs). For example, the research highlighted how wider socio-economic and political issues linked to the impact of the cost of living crisis on already impoverished families, many of whom were already affected by restrictive changes such as capping of child benefits (two-child rule) was a key driver of food insecurity. Importantly, this work identified complex and often precarious interdependencies between CFOs, the wider food sector (i.e. surplus food) and local authorities. Given the concurrent cost-of-living crisis and changing political agendas, providing an understanding of the fragility of the system (through policy briefs)⁽⁷⁵⁾ to local and national governments has been important and timely to support decisions about future funding models.

Systems approaches to develop interventions

Systems thinking can be drawn upon to support development of interventions that offer greatest potential for disrupting the food system. This begins with adopting the basic principle of hearing from multiple perspectives, to understand how a complex system functions, prior to developing the intervention^{(Reference Hammond76,Reference Savona, Brown and Macauley77)} . This includes defining the problem that the intervention aims to solve, and constructing a visual representation of contributing factors, as well as the connections and relationships between those factors^{(Reference Siokou, Morgan and Shiell78)}. Once an understanding of the system has been constructed, ‘leverage points’ can then be identified by intervention developers. These are places within a system that, if changed, could lead to a large shift in how that system functions^{(Reference Meadows63)}; for example, adjusting collective goals that parts of the system are working towards, or altering the mindset of individuals who have the power to change parts of a system. Donella Meadows developed a list of ‘12 places to intervene in complex systems’ in 2008^{(Reference Meadows3)}, which was intended to be tentative and was designed to invite intervention developers to think strategically about which leverage points could be influenced to achieve maximum disruption, and consider the direction in which to ‘push on them’. Since then, others have adapted this list of ‘places to intervene in a complex system’ to create more accessible and reproducible leverage point frameworks to be used within a public health context, including the Intervention Level framework, which collapses the list of twelve places to intervene into five levels of intervention^{(Reference Johnston, Matteson and Finegood79)}. The Action Scales Model^{(Reference Nobles, Radley and Mytton80)} also aligns with Meadow’s twelve places to intervene, but conceptualises places to intervene/leverage points across four levels. More recently, Bolton et al.’s Public Health 12 Framework^{(Reference Bolton, Whelan and Fraser81)}, aligns to Meadow’s twelve places to intervene, but has adapted the language to better meet the needs of public health practitioners.

One example of where intervention development considered system leverage points includes the CONNECTS-Food resource, an intervention which aims to promote the adoption of a ‘whole school approach’ to food in UK primary schools^{(82,Reference Bryant, Burton and O’Kane83)} . Research began by using a group model approach to consider (and map) factors that influence children’s food choice throughout the school day. School stakeholders with differing perspectives played a key role in this and workshops with them were conducted alongside journey mapping workshops with children. A network analysis was then used to describe the relationships and interactions between factors within the system, identifying 27 thematic nodes contained within four overarching themes (Figure 1). Using this map, researchers were able to identify internal and external factors that influence school food, ideal leverage points for intervention, and smaller points of influence that may be more feasible to intervene at given resource constraints; for example, influencing school leadership to prioritise food was identified as a clear leverage point. Consequently, CONNECTS-Food resources provide school stakeholders with practical recommendations to enable food prioritisation by school leaders, whilst also aligning to wider influences such as food policy, school governor priorities and Department for Education monitoring. In turn, the map tells us that achieving this goal has the potential to influence wider parts of the system through the relationship between leadership and other key factors, such as the school food environment and food curriculum.

Figure 1

School Food System, provided with permission from Bryant et al., IJBNPA; 2023;20(1):1–11^{(Reference Bryant, Burton and O’Kane83)}.

The importance of food system transformation more widely is recognised by government leaders and has resulted in a number of research funding calls to address multiple aspects of the food system, including a cross-funder Transforming UK Food Systems (TUKFS) programme⁽⁸⁴⁾. The TUKFS-funded FixOurFood project^{(Reference Doherty, Bryant and Denby41)} used a systems-informed Three Horizons futures process to identify priority domains of transformative action for three individual sub-systems of the Yorkshire food system (agriculture, schools and early years settings and food economies), and also collectively for the whole food system (Figure 2)^{(Reference Buckton, Fazey and Doherty67)}. In this case, priority action domains were identified based on an understanding of: the fundamental contrast between the current food system and a regenerative desired future food system; what is needed for genuinely disruptive and transformative action (‘H2+’) as opposed to action that simply props up current failing systems (‘H2-’); how the action domains would reinforce each other; and where momentum for action already existed in the food system^{(Reference Buckton, Fazey and Doherty67)}. The six resulting actions include: (1) enhancing supply chain connectivity and innovation to support diverse hybrid business ecosystems; (2) scaling environmentally beneficial and regenerative farming; (3) empowering consumers to reshape food demand; (4) providing trusted, accessible knowledge support for standards and incentives; (5) supporting schools and young people as drivers of long-term change; and (6) ‘cohere, integrate and coordinate’ (ensuring coordination and mutual support across domains)^{(Reference Buckton, Fazey and Doherty67)}.

Figure 2

FixOurFood’s Three Horizons map of current Yorkshire food system challenges (Horizon 1, H1), the desired regenerative future food system (Horizon 3, H3), and key domains of action for supporting the transformation from H1 to H3 (Horizon 2, H2)^{(Reference Buckton, Fazey and Sharpe66)}. The scaling used in the graph is not intended to be taken literally, but rather for rough qualitative comparison. ‘Prevalence’ describes the relative prevalence of a horizon in the food system (e.g. the amount of resources it holds or transmits, or the proportion of human activity taking place that is related to the horizon). Time flows from left to right: the present is found where H1 is dominant, and all information to the right of this is in the future. The exact placement of the circles vertically and horizontally is not intended to reflect differences in prevalence or time – the important aspect is the circles’ association to a particular horizon.

Systems evaluation approaches

There may be tensions between those who argue for evaluation approaches to define effectiveness using comparator designs and those who favour approaches that understand the potential for impact^{(Reference Paparini, Green and Papoutsi85–Reference Attanasio and Cavatorta87)}. However, application of systems thinking for evaluation should not be considered as a replacement for randomised controlled trials. Rather, systems evaluation provides the potential to help understand how, and why an intervention may or not be deemed effective^{(Reference Moore, Evans and Hawkins88)}. This extends more traditional process evaluation asking about the population reach and fidelity of interventions and moves towards understanding context; exploring why intervention goals may or may not have been successful and how relationships between components or approaches can be enhanced to support implementation and thereby increase effectiveness. Again, any tensions might be diffused simply by aligning the most appropriate evaluation approach to answer the research question. Thus, if an intervention plans to operate across a whole system to improve outcomes, systems evaluation will facilitate an understanding of where the boundaries lie (and where any effects are most likely to happen), capture information about the pre-existing system, monitoring dynamic programme outputs and considering these alongside any outcomes. This approach has been formalised into frameworks, including the ‘Evaluation of Programmes in Complex Adaptive Systems’ (ENCOMPASS) framework for public health interventions^{(Reference Pinzon, Stronks and Dijkstra89)}. ENCOMPASS is being applied to support the evaluation of the whole-city approach to obesity in Amsterdam^{(Reference Sawyer, den Hertog and Verhoeff90)}, Bradford’s Good Food Strategy⁽⁹¹⁾, and the large UK Prevention Research Programme, ActEarly^{(Reference Wright, Hayward and West92)} which aims to create ‘City Collaboratory’ testbeds to support the identification, implementation and evaluation of upstream interventions within a whole-system city setting to improve child health outcomes. Rather than solely focusing on inputs and outcomes, the ActEarly evaluation seeks to explore changes across various processes of researcher collaboration, local authority partnership working, co-production, citizen science and data linkage, whereby ActEarly is a type of ‘meta-system’ in which wider policy decisions are made to direct resources to different, and often competing, causes^{(Reference Mansukoski, Lockyer and Creaser93,Reference Nixon, Sheard and Sheringham94)} . This approach does not seek to understand intervention effectiveness using comparative approaches, but rather to map and explore the context, mechanisms and outcomes of implementation of a collective, interlinked series of large system change interventions in order to estimate whether such implementation is able to enact systems change towards a tipping point (where ActEarly processes/interventions /collaborations continue to function independently of the ActEarly programme). Again, such approaches are not new, but are often considered innovative with the context of public health and local food systems. FixOurFood is using a similar, co-designed, approach to evaluate a new local government (North Yorkshire Council) food strategy, drawing on recent work synthesising principles for transformation-focused evaluation^{(Reference Buckton, Fazey and Ball95)}. Various other systems-informed evaluation frameworks exist^{(Reference Buckton, Fazey and Ball95)}, such as Margaret Hargreaves’ framework for evaluating system change^{(Reference Hargreaves96)}.

Elicitation of perspectives and systems governance

At the heart of any systems research, is the need to elicit perspectives from key stakeholders and beneficiaries. It is through the eyes of those who ‘live’ and experience the system that we gain a clearer understanding of what the system looks like and how its parts interact. This extends the concept of food systems governance from one where decision-makers influence processes, such as policy, management, organisation, measurement and coordination^{(Reference Yap97)} to include decision-making at the consumer and beneficiary level. A number of methods are used to capture insights, with ‘group model building’ being one of the most common^{(Reference Siokou, Morgan and Shiell78,Reference Doyle and Ford98)} . This participatory process begins with a facilitated group discussion to better understand the issues or factors linked to problems within the system aimed at gaining a consensus view, not just about what the system looks like, but also a dynamic hypothesis about how the system’s behaviour changes over time. Importantly, the decision of who to invite to provide stakeholder perspectives is crucial to understanding factors that cover the breadth and factors within the boundaries of the system under investigation. In recent work, a group model-building approach was used to explore factors that people felt influenced food intake in children aged 0–5 years, and thus, develop an understanding of the obesity system for young children. This process engaged with a large group of early years practitioners, parents and national experts^{(Reference Bryant, Burton and Collinson99)}. The resulting ‘early years obesity systems map’ includes some of the factors previously identified by the (adult) Foresight obesity map, but with additional factors linked to the role that parents and families play in food choice during the early years, in addition to the role of early years settings (such as childcare, nurseries and the extended family). However, given the nature of the stakeholders, there were gaps in factors that are apparent in the Foresight map, most notably, biological drivers of appetite regulation. As such, a decision was made to complement the perspectives of the stakeholders with a brief review of the evidence base. In this example, we demonstrate that the approach is not rigid, but that all efforts should be taken to carefully consider the range of relevant stakeholders that are invited to take part in the process.

Critique of systems thinking approaches

Systems thinking approaches, at least in Western contexts, have positivist philosophical foundations^{(Reference Morcol100)}. A positivist approach to systems would assume that given a complete understanding of a system, all its parts and their relationships, the system’s behaviour could be understood and predicted^{(Reference Levy101)}. Thus, a system model that failed to match (or potentially) predict the real system’s behaviour, would reflect the modeller’s failure to fully understand and map that system. It was also thought that increased system complexity was a consequence of the size of a system, the number of parts and the relationships between those parts.

Complexity theory, which emerged at least partly out of the work of second-order cybernetics such as Von Foerster^{(Reference Von Foerster102)}, Mead^{(Reference Mead4)} and Basteson^{(Reference Bateson103)}, and developed further by Kauffman^{(Reference Kauffman104)}, took a different view. Complex systems demonstrate behaviours that cannot be predicted solely from knowledge of the system’s parts and their relationships. Examples of such behaviours are self-organisation, emergence, nonlinearity and inertia. Nor do any individual parts of a system exhibit those behaviours on their own; hence ‘the whole is greater than the sum of its parts’. This presents a significant challenge to those wishing to model and intervene in complex systems as precise interventions are impossible and changes in system behaviour difficult to causally attribute. Behaviours such as nonlinearity (where observed changes within the system are not proportional to the level of input) and inertia (resisting change due to competing pressures) also mean that there may be a significant delay in the response of a system to any intervention (which can be a significant problem for academic and political timescales). One outcome of this is that we perhaps do still see a tendency by some academic researchers, and attempted interventions in complex systems, to revert to more classical systems theory and not adequately accounting for systems’ complexity. The result of which could be claims about causality that are not well supported^{(Reference Crielaard, Quax and Sawyer105)}.

These issues are perhaps at their core due to classical systems approaches being inherently reductionist (perhaps due to an underlying fear of uncertainty and loss of control). Modellers extract parts of a system, and some of the relationships between those parts, into a simplified model in the hope that it will be good enough to enable more considered interventions with higher chance of success (as outlined above), and make a problem more tractable. This is not inherently bad, as long as we do not lose sight of the limitations of these approaches, and fall into the trap of drawing simplistic causal interferences between intervention and outcome. Systems approaches might instead be used in combinations as exploratory tools in food and health systems research, whilst also adopting complex thinking approaches^{(Reference Teixeira de Melo, Caves and Dewitt106,Reference Morin107)} .

Complex thinking has, at its core, the notion that the process by which you think about a problem should mirror the complexity of the problem. Complex thinking, like cybernetics, is therefore not reductionist. For example, Ashby’s law of requisite variety broadly states that the variety (similar to complexity) of the system being used to control a complex system must have enough complexity itself, or it will be ‘dominated and ultimately destroyed’ by the thing it is trying to control^{(Reference Ashby108)}. Or to put it another way, it must have requisite variety. Thus, using a relatively simple systems map of something very complex to govern that system, and make interventions, might not work on its own. To put this into the context of food or health systems, Ashby’s law would suggest that the infrastructure that we build to govern food and health systems must itself be of sufficient complexity, or it will be destroyed by those systems, which is arguably what we are observing today.