Reconsidering the Role of the Digital in Global Production Networks

Global production networks (GPN) has become a key framework in conceptualising linkages, power and structure in globalised production. However, this framework has been less successful in integrating the influence of digital information and ICTs in production, and this problematic in a world where relationships and power are increasingly mediated by digital information flows and resources. We thus look to adapt the GPN framework to allow more substantive analysis of ‘the digital’. Primarily this is done through a theoretical analysis of the three core categories of the GPN framework – embeddedness, value and networks – to highlight how these categories can better integrate a more dynamic and contested conceptualisation of the digital. Illustrations from research on the digitalization of tea sector GPNs in East Africa highlight how these theoretical advances provide new insights on the digital and its expanding role in economic production.


Introduction
Global production networks (GPN) has become a key framework in conceptualising linkages and crucially uneven power and structure in globalised production (Coe & Yeung 2015, Henderson et al. 2002. The GPN framework explores the economic relations of actors directly involved in economic production, as well as a wider gamut of networked relationships, institutions and policy that orientate the forms and practices of production. This paper originates from research exploring changing practices around digital information flows, digital data and information & communication technologies (ICTs) in production -we will refer to this socio-technical bundle of activities as 'the digital' in this paper (as outlined in the next section). In previous research we explored the impacts that the digital might play in reconfiguring production, with a particular interest on low-income producers in the Global South.
Initially, the GPN framework appeared to be an appropriate choice of framework. It would allow us to conceptualise the broader relationships, and power related to the producers we were examining, and it would form a basis for understanding how the digital might disrupt economic structures and relationships. Yet, in practice, we found that the GPN framework in its current form provided only limited analytical value in our investigations.
As has been detailed elsewhere (Foster & Graham 2015b, Foster & Graham 2015c, Waema & Katua 2014, our empirical research findings have suggested that the digital is increasingly contested element of production. Active exclusionary practices have led to overt contestation around access to digital information flows, and services (e.g. where agri-business actors make it difficult for smallholder farmers to access information resources), whilst less visible affordances and functioning of ICTs and digital software subtly re-enforce the disadvantages that small firms have (e.g. global tourism services where requirements for digitalization often exclude small firms). These findings, that the digital becomes a 'site' of contestation and uneven relations, were difficult to integrate into core usage of the GPN framework and thus for all its explanatory potential, we found limited guidance on integrating the digital resources as an important role into GPN analysis. This paper argues that the poor explanatory power of the GPN framework to understand the digital is more than an isolated issue for our specific research. We found limited conceptual literature on this topic, and whilst literature does often acknowledge that the digital plays an important role in tying together increasingly geographically dispersed production networks, it rarely explores the digital as anything more than an infrastructural component. The relative absence of research contrasts with the fact that digital data flows and services are increasingly central to the efficient management and monitoring of global dispersed networks. Innovation and competitive advantage in production is increasingly associated with the rollout and integration of digital services, digital networks and online applications (Malecki & Moriset 2007). Therefore, it is important to critically analyse the role of the digital in shaping global production. This paper thus seeks to better understand how to embed the digital within thinking about GPNs and poses the following research question: How can the GPN framework best capture the important roles that the digital plays in shaping GPNs?
The paper is set out as follows: it first introduces the ways that the GPN literature has discussed ICTs and digital flows of information and defines the terminology of 'the digital'. It argues that there are components, already present in the GPN framework that can be explored to help reconceptualise the role of the digital in more fruitful ways. Second, following on from this argument, we draw on a literature analysis to highlight how the three key conceptual categories of GPNs -embeddedness, value and power -can better integrate the digital. Third, to expand understanding and demonstrate how these re-conceptualisations can be employed in practice, we turn to an empirical case study of the role of the digital in networks of Rwandan tea production.

'The digital' and GPN frameworks
With the world's growing access to ICTs, and increasing codification and digitalization of production, digital data, assets and information flows have become a key element of production across all economic sectors. A growth of digital information flows (whether that be the digitization of previously analogue information, the codification of new information, or the emergence of new digital services) is driven by the demands of more granular financial and managerial control, the agility to improve production and demands for more rapid innovation (ibid.). Digital assets and information flows can play many roles, but frequently supports interfirm production relationships, planning and co-ordination, and enables more granular and rapid knowledge of production processes, and customer behaviours (Castells 2000, Chen 2002). Thus, as digital resources and ICTs become more central to production, we need to better understand their impacts on production networks. The goal of this paper is to fill this gap and to provide clearer theoretical directions as a first step to systematically analysing global production in an increasingly digital world.
We use of the notion of 'the digital' in this paper to define the constellation of digital ICTs, connectivity, infrastructure, data and new digital resources which are increasingly crucial in all aspects of the economy. This idea of 'the digital' is built around Berry's (2015) call for a 'post-digital' analysis.
That is, an analysis where ICT, information flows and data are increasingly not a 'virtual' add-on to the real world but entwined in everyday economic activities. He notes: "the world is transitioning from analogue, structured in most part by the physicality of destination, to the digital. A new industrial internet is emerging, a computational, real-time streaming ecology that is reconfigured in terms of digital flows, fluidities and movement. (p2) …this also signals a move away from a previous 'digital' era that was tangential to the capitalist economy, but nonetheless facilitated many economic growth regimes associated with it, such as ICT, finance-led and so forth. Instead we are entering a post-digital world in which the digital has become completely bound up with and constitutive of everyday life and the so-called digital economy." (p3) As Berry notes, analysis of the digital needs to move beyond rather static analysis of digital ICTs and digitisation of certain flows. Rather we are interested in making a wider socio-technical analysis that emerges from new digital technologies, connectivity, services and ICTs; and the increasingly intense and dynamic information flows and the new assets they facilitate.
Whilst the digital has not been part of the core focus of GPN analysis, there are many mentions of the importance of the digital in supporting the globalisation of production (Dicken 2011). More specifically as theorised within the seminal paper which outlines the GPN framework, ICTs are discussed as a cross-cutting element.
"We exclude 'technology' as a separate category. Instead, ICT is rather seen as an inherent element of GPNs, underlying the development and maintenance of network connections. Technology, as one of globalization's drivers, influences the processes of value creation in different places, as well as transforming the means by which power is exercised. Additionally, it affects the agents' possibilities of embedding in, and disembedding from, particular networks and territories." (Henderson et al. 2002 p.447) This quote hints at a variety of directions for considering the role of the digital -there are potential directions to exploring the links between the digital and infrastructure, value and power that might be expanded. GPN approaches often focus on the concept of embeddedness -that is, understanding the coupling between networks of global production and the diverse locations within which production networks are grounded (Coe et al. 2008, Coe et al. 2004, Hess 2004. As detailed in the next section, the focus on embeddedness has often led to the digital becoming a background element, an infrastructure which supports non-territorial forms of network embeddedness. The paper argues that whilst embeddedness can provide insights into the digital (e.g. Wrigley & Currah 2006), it does not have sufficient breadth to fully represent the growing and dynamic role that the digital is playing in production, in economic relations of firms in GPNs, and ultimately in development.
Using the GPN framework offers significant advances for explaining aspects of firm relations, loci of power and geographies of production, and provides a sophisticated understanding of the global fragmentation of work. Thus, finding richer ways of conceptualising the digital in the GPN framework is liable to feed into better understandings of relationships and power, and consequently the formulation of more contextually appropriate policy around global production.

Exploring GPN Models
To build a clearer understanding of the digital in GPNs, we explore the three main 'conceptual categories' of the GPN framework -embeddedness, value and power. The success of the GPN framework owes much to the way that the conceptual categories have been coherently integrated in the framework. At the same time, they sit uneasily and at times draw on diverse ontologies (Sunley 2008).
By detailing these diverse ideas, our goal is to consider the different components that make up GPNs.
We will later use these components to build a framework to systematically explore the effects of digitalization in GPNs.
We particularly draw here on early GPN literature that introduces key concepts and categories, work which still provides the most substantial theorisation of the GPN framework (Dicken et al. 2001, Henderson et al. 2002, Hess 2004. Recent academic contributions whilst preserving underlying concepts has particularly looked to focus on the causal drivers of production networks (Coe & Yeung 2015). For instance, cost reduction, firm capabilities, labour power, firm entry barriers and financial discipline have been articulated as important shapers of firm strategies that influence production networks (Barrientos et al. 2011, Coe & Yeung 2015, Mahutga 2012). Our approach is complimentary, in that studies of the digital explore ways by which these 'causal drivers' are operationalised in global production networks (i.e. the ways that digital technologies are increasingly important in supporting cost reduction, financial oversight, restructuring of labour etc. in GPNs). However, in the spirit of older GPN work, we also will argue for making consideration of less deterministic viewpoints. Active strategies can be negotiated, resisted or avoided (Ouma 2010), and in terms of the digital this can be in unexpected spillover effects, path dependency of digital artefacts and appropriation: which means that actual outcomes often differ significantly from those intended by economic actors with agency.

Embeddedness
A core element of the GPN framework is the 'embeddedness' of economic relations. The utility of 'embeddedness,' is that it allows for an exploration of the ways that actors are orientated by socioinstitutional influences through relations within localities, production networks or from firm origins.
As first outlined in Granovetter's seminal paper on the topic, economic embeddedness suggests that underlying all economic transactions are a number of social and cultural norms (Granovetter 1985).
Embeddedness thus grounds economic relations within socio-cultural factors rather than them occurring in some abstract value-free economic space.
Earlier analysis of embeddedness in economic geography has principally explored how local economic exchanges were defined by the trust, norms and rules, constituted at a local or national level. The GPN framework goes beyond a one-dimensional perspective of embeddedness to explore a variety of other forms of embedded exchanges as outlined by Hess(2004).
"economic geography has been prone to use what I will call an 'overterritorialized' concept of embeddedness by proposing 'local' networks and localized social relationships as the spatial, logic of embeddedness, which might result from 'spatial Fetishization' " (ibid. p.174) Drawing on such arguments, the GPN framework expands the definitions of embeddedness to consider what has been referred to as 'network embeddedness', that is the "connections between network members regardless of their country of origin or local anchoring in particular places. It is most notably the 'architecture', durability and stability of these relations, both formal and informal, which determines the agents' individual network embeddedness" (Henderson et al. 2002 p.443) Thus, network embeddedness can be thought of as a gamut of sectoral norms, rules, professions and institutions that orientate the activities of firms in global production. An additional category of societal embeddedness is also sometimes used 1 , which refers to the culture that firms involved in global production bring with them from their home country, which continue to influence as they expand globally (Henderson et al. 2002, Hess 2004). The justification for use of the two/three categories of embeddedness and their differentiation (territorial, network/societal) is never fully justified, but there is a nod to Castells and his concept of 'space of places' and 'space of flows' to explore the coupling between durable territorial norms and those in network relations (Castells 2000, Henderson et al. 2002. As mentioned in the previous section, the digital is only elusively present in theorisations of embeddedness. The digital tends to be framed as background infrastructural elements that facilitate network embeddedness. Digital information flows and ICTs are barely mentioned in name, but it can be inferred that the durability of non-territorialised relations (and consequent embedded exchange) come at least in part from digital infrastructure and networks that enable the durable exchanges in GPNs.

Value
The notion of 'value' is used to outline the actors and positions within networks where rents (frequently economic ones) are created, enhanced or captured. Two aspects of value are particularly influenced by the digital. First the introduction and expansion of new digital information flows, services and networks which may alter the economic value distribution within production networks. Second, we are seeing new types of 'digital production network' emerging, in areas like online services, knowledge work and the web production (Beerepoot & Lambregts 2015, Lehdonvirta et al. 2015. Often 'digital production networks' are still at nascent states, and there is significant research to be done to understand the concept of value in such networks (Löfgren 2003, Terranova 2000. We therefore focus only on the former category in this paper, a focus that will be most relevant for those exploring the digital in existing GPNs. In terms of value, GPN models often closely align with global commodity chain (GCC) and global value chain (GVC) approaches (Coe & Yeung 2015). In particular, chain based models have been effective in exploring the distribution and governance of value in interconnected networks of production. In earlier commodity chain literature, the notion of value was more closely aligned with surplus value that emerged out of differences between labour costs of commodity production and the exchange value of those goods. Forms of network governance (and consequently the distribution of value in networks) were articulated to emerge mainly as a result of innate characteristics of a particular production sector -that is to say that value distribution principally emerged from the products, production requirements or sectoral structures within which production existed (Gereffi 1994). More recent value chain research has probed networks at a more granular level, focussing more on aspects of value which emerge from economic rents, and those activities such as innovation, branding and business models that allow firms to charge economic premium on goods and services. Value chain models highlight the increasingly sophisticated ways that production is organised in networks (such as by modularisation of production and standardisation), even as the products and processes in these chains have become ever more complex and personalised (Gereffi et al. 2005, Sturgeon 2002). This more recent work has argued that value distribution is not only an outcome of sectoral properties and labour, but can be enhanced as value chains are designed, managed and improved with appropriate modularisation, quality control or institutional oversight (Kaplinsky & Morris 2001). This more recent value chain research then opens up a more purposive way of thinking about value, where ICTs and digital information flows can be one way of reconfiguring and shaping network relations and value distribution.
There have been a number of studies that have explored the effects that the digital can play in supporting production structures and exchange. Notably, work during the emergence of the Internet discussed the potential for digital networks to rework production processes and chains (such as Benjamin & Wigand 1995, Gereffi 2001. In particular, Gereffi outlined some of the implications of the Internet that might transform network governance: new online channels might allow producers to capture more value at the expense of intermediaries; interfirm digital services could reduce chain management and transaction costs; and the ability for personalisation could lead to profits becoming centred on the ability for customer personalisation. However, this literature drew on limited empirical evidence from a time when the digital services and ICT were still diffusing and tended to reflect optimistic visions of new digital networks. We argue that there are also some theoretical weaknesses in the way that value and the digital have been considered. There is a tendency to dwell on value creation and enhancement over value capture -that is there is more discussion of the new products, services or valueaddition that are driven by the digital, but there is less attention on who (and where) such the surpluses accrue to (Coe & Yeung 2015). There is also a risk that research on the digital and GPN focusses on value only economic rents and the ways in which the digital transforms business without understanding how digital technology may also impact on the ways of producing surplus value. It is well known that digital technologies not only transform business but in the process transform work patterns and divisions of labour (Castells 2000) and it is important to consider these impacts is analysis of value.
Thus, we only have a partial picture of the impacts of digital technologies on the distribution of value.
Will the promised improvement be accessible to all actors in GPNs? Where in production networks will the 'value' from the digital be captured? How will the digital also impact the division of labour and structures of work?

Power
In Henderson et al's. seminal work outlining the GPN framework, power is seen to be exercised in three different ways: corporate (i.e. firms), institutional (i.e. state, global institutions) and collective (i.e. unions, NGOs) (Henderson et al. 2002), all which imply wilful agent-led activities with little room to conceptualise the digital. However, an analysis of how actor-network theory (ANT) has been integrated in the GPN framework can serve as a starting point to think about the power and the digital.
The GPN framework is strongly influenced by ANT thinking of networks constantly in the making, and ANT ideas strongly cohere with case studies of global production networks as often complex and reconfigurable (Chen 2002, Lüthje 2002. The ANT notion of the duality of actors and networks is key, it problematizes actors, who from ANT perspectives are considered to be a constituted by networks of other actors (Latour 1999). Duality is powerful in that it provides a theoretical basis for the GPN framework to move beyond linear explorations of production relations as exemplified by GCC/GVC approaches. If each actor in the value chain is constituted by a whole gamut of networks relations and actors, one also needs to consider these constituting actor-networks (Coe et al. 2008). Thus ANT perspectives highlight more distributed sources of power where global production is shaped by interactions within localities and networks, and allows traversing between different scales of economic interactions.
However, without the inclusion of the 'non-human' as an actor with agency, the outline sketched out above only presents a partial picture of ANT and power. Inclusion of non-human actors prompts us to consider the role that objectified resources (including digital objects and information flows) play in influencing networks, and how these non-humans influence activity (Latour 2005). GPN literature has made reference to non-human actors in relation to power previously, particularly in early theoretical work (Dicken et al. 2001, Yeung 2003, "The implication [of ANT] for understanding the development of the global economy is that analysts need to recognize 'agency' as an 'effect generated by a network of heterogeneous, interacting materials' in which the nonhuman plays a critical role in embodying and shaping (often in unexpected ways) action" (Dicken et al. 2001 p.102) However, there appears to be some friction amongst key architects of the GPN framework. For instance Coe & Yeung(2015) have argued that one of the key differentiators of the GPN framework from simply using ANT on its own, is that the notion of non-human 'agency' should not be integrated within GPN notions of agency and power 2 .
"To be clear, GPN1.0 does not adopt ANT wholesale. It does not, for example, confer agency upon non-human objects as ANT does" (ibid. p.12) Whilst we are sympathetic that the notion of 'agency' may be a problematic terminology to explain the role of objects in guiding and constrain networks, this should not lead to a partial use of ANT in GPN analysis. Marginalisation of ANT is detrimental to understanding how the digital relates to power and as shown by growing use of ANT in neighbouring fields such as information systems and organisational studies, ANT can provide new insights on organisation (e.g. Ciborra et al. 2000, Orlikowski 2010).
Literature in these neighbouring fields have explored digital technologies through the entanglements of objectified structure and social structure, they have thereby attributed a more active role to ICTs, where affordances, inscriptions and constraints of technology can guide, define or restrain human agency and relationships. These literatures point towards approaches to power that the GPN framework could also look to integrate.

Summary
The three conceptual categories and the ways that the digital has been integrated are shown in Table 1.
'Embeddedness' examines the durable relations and norms that guide economic activity, which have their basis in both localities and networks. 'Value' explores the ways that value is created, enhanced and distributed in networks of firms, and how distribution of value can be improved or changed. 'Power' explores the shifting nature of control in networks, and the multiple scales at which networks act. Within all three conceptual categories, we have highlighted deficiencies in conceptualising the digital.
The deficiencies are similar in all three categories: the digital is rarely problematized as a changing, dynamic and active element in globalised economy. It is either treated as a background element or ignored. The literature review has alluded to some ways in which the digital might be better brought into GPN analysis -as a more foregrounded element in embeddedness, treated more critically in the distribution of value, and as a non-human actant with effects on power structures. Yet these insights are not complete and we need clearer illustrations to better guide how they allow us to better analyse the digital.

Approach
We present a set of illustrations, as extracts, to expand on how we can better integrate the digital into the core categories of the GPN framework. The goal here is add some clear empirical examples of digital activity to support the substantial theoretical discussion, and to help expand on some of the key insights made.
To explore links between GPNs and the digital, we discuss research undertaken in the GPNs of tea production with a particular focus on Rwandan production. Our case selection here originates from our interest in the integration of digital technologies closer to producers, particularly more marginal groups in the Global South. In contexts of the digital, this case also lucidly illustrates that even in longestablished commodity sectors largely characterised by material goods, we are seeing the growing importance of 'the digital' in reshaping the sector.
In recent years, Rwanda has been undergoing significant transformation in its digital connectivity, with the construction of a series of submarine fibre-optic cables linking the region to the rest of the world (before 2009, the region was entirely dependent on satellite connectivity) (Foster & Graham 2015a).
Additionally, the growth of ICT and mobile phone use in the region means that even low income producers may be able to access digital information flows (Foster & Heeks 2013). The goal here is to illustrate some of the impacts of the digital on tea production networks: particularly focusing on how the digital might impact on how Rwandan actors link into these networks using the core categories discussed previously.

Introducing Rwandan tea production
Rwandan tea farmers have becoming increasingly interconnected into the global production of tea in the last decade. This change has notably occurred as part of wider trends of developing country tea production moving away from state control in government tea boards towards private sector development. Consequently tea operations in Rwanda such as tea processing 3 and transportation are now run by private firms, including entrance of large tea multinationals buy-outs, overseen by a commercially orientated regulator -the National Agricultural Export Board (NAEB).
Privatisation in Rwanda has been strongly supported by political leadership in Rwanda, who are pushing economic growth through private sector development, a drive which has led to the country being dubbed 'Rwanda Inc.' driven by president-cum-CEO Kagame (Crisafulli & Redmond 2012).
Nevertheless, it is still necessary to ensure stability in the country, given that Rwanda continues to bear the scars of recent conflicts. In industry, ensuring stability results in political actors having control of key firms in the economy, what Booth & Golooba-Mutebi(2012) have referred to as 'developmental patrimonalism'. Thus, even as the tea sector moves towards privatisation, the state still exercises power, mainly indirectly through its influence on key firms and regulators.

Figure 1: Stylised outline of operational links linked to Rwandan tea production Source: Authors fieldwork
As shown in Figure 1, tea production in Rwanda is well integrated into global networks of tea distribution. Green tea leaves produced mainly by rural smallholders is sold via cooperative associations to tea factories which undertake preliminary processing. Processed tea then passes through a set of intermediary trading stages (which will be discussed below) which results in tea being bought in Kenya, packaged and transported to international markets by large tea firms. Some Rwandan tea is sold directly to retailers (as shown in the middle of the diagram), but the majority still tends to be sold via an intermediary auction in Mombasa, Kenya.
As with other agricultural sectors, tea is moving away from being a generic commodity, and in Rwanda varying types and marks of tea are targeting a range of customers: some tea producers are contributing to high value-added, high quality teas bought by boutique tea firms abroad, but the majority is still bought by large tea retailers and used as part of tea blends 4 for commercial retailing. An increasing volume is also sold (as we show in the bottom-right of the diagram) into regional or emerging market consumers in locations such as Pakistan, Central Asia and the Middle-East (Africa Tea Brokers 2015).
A key finding of our research in Rwanda was that private sector development had led to only limited modernisation of tea processing machinery or improved tea farming practices. However, the growth and introduction of information technologies and digital connectivity has been significant. All tea processors interviewed had internet connectivity: a surprising fact given the remote location of some of the facilities requiring, in some cases, construction of towers within their grounds to link to wireless internet 5 . Equally, many elements of the production process, such as data on green tea quantities from farmers, processed tea batches, and tea quality, have been codified, digitized and shared to allow improved management and traceability of tea goods as discussed below.

Embeddedness
In the Rwandan tea sector we explored how tea firms were integrating into international tea production networks (and consequently the role that the digital was playing in influencing the norms, rules and institutions in Rwanda). A key finding in this analysis was that the integration of Rwandan firms into international production did not entail a consistent set of requirements or influences locally. Rather Rwandan firms were making diverse, and multiple different economic linkages each with their own characteristics of activity and requirements.
Two illustrative network linkages in Rwandan tea highlight these contrasting norms and requirements of different networks. One was around auction selling of tea (the upper path from left to right in Figure   1), the other around direct selling of tea (the path in Figure 1 passing through the 'direct buyer'). These different networks had distinctive institutions and approaches to trust. Auction selling is a many-to-one exchange from producer to retailer where economic relations revolve around the rules, norms, and institutions that have emerged over a long time in the region to ensure trust in auction exchanges.
Direct selling networks tend to be a one-to-one and are grounded in direct high trust firm-to-firm relations. Tea processors in Rwanda may build relations in one or both of these networks (dependent on tea quality, and demand), and this can lead to contradictions. For example, direct selling networks push integrated economic relationships that may reduce local firms and policymakers ability to dictate production (e.g. less transparent pricing, less opacity in exchanges). In contrast, as firms in Rwanda have looked to value-addition of tea (fair trade, packaged or new types of tea), the norms of the auction network have been slow to adapt to these changes (e.g. respondents commented on difficulty in integrating certifications into auction norms).

Figure 2: Simplified and generalised schematic of key digital data flows in the production of Rwandan tea Source: Authors' fieldwork
The digital also plays more than just an underlying infrastructural role as shown in Figure 2. Auction selling was characterised by 'digital fragmentation', fragmented and actor-initiated digital exchanges (e.g. through email exchanges, payment services, looking up online catalogues, spreadsheet updating). In contrast direct selling was characterised by 'digital integration', the automation of a whole slew of digital parameters as part of integrated tracking and monitoring of production, transport and goods. These two modes of digitally mediated exchange underlie the economic exchange and had significant effects on the way specific Rwandan firms were interacting. Digital fragmentation allowed firms in Rwanda more leeway to integrate the requirements of international production within local norms. On the other hand digital integration acted in many ways like a bulldozer. Some Rwandan tea producers were in the process of integrating with software and systems defined by parent tea companies and the requirements of these systems were beginning to orientate many new element of activity reducing the ability for any local adaptation of activities.
In sum, this illustration highlights that the GPN concept of network embeddedness -that set of durable interactions that are not territorial mediated -are not necessarily singularly coherent. Expanding connectivity enables a diverse range of networked interactions that are liable to lead to a range of different durable network relations. Moreover, the digital infrastructure itself which facilitated network interactions was important, in the ways that platforms and technology orientated the types of norms and activities that define embeddedness.

Value
As has been outlined previously, achieving standards and sufficient quality of goods is an increasingly important element of agricultural production, and is entwined with the economic value of those goods (Gwynne 2006, Ouma 2010. In Rwandan tea, one way in which standardisation and quality is pushed is that exchange was increasingly mediated via digital technologies as a means to allow exchange to be monitored. Rwandan tea processors now operate 'weighbridges' that enable them to verify and digitally integrate the weight of green tea leaves which arrive at the processing facilities from farmer co-operatives (groups of smallholder tea farmers use co-operatives associations to simplify transport, payment etc).
Some tea processing plants have also adopted more advanced field weighing where weight of harvested tea leaves from smallholders are integrated into tea processor information systems.
Intuitively, such digital devices, which simplify the complexity of green leaf weighing and payments, appear to offer reduced corruption and modernisation of tea production. For instance, with field weighing, one large tea cooperative association's chairman whose farmers had recently begun to use these systems outlined the benefits:

"With the old system, we used to record slight differences in what has been weighed in the fields and what has been weighed at the factory [tea processor] but today it is all the same. The new system is much better and the beneficiaries never stop commending its benefits."
However, digging deeper, one increasingly sees risks with this technology. The new digital systems benefits some, but for others they were exclusionary. For example, one leader of a smaller cooperative association made up of very low income farmers described the impact that digital weighing Thus, technology mediated exchange has not benefitted all stakeholders equally. It may have a longerterm exclusionary effect on certain smallholder farmers by excluding tea cooperatives associations. This is important as co-operative associations have traditionally been an important source of advocacy for the rights of smaller growers and pluckers in the tea sector.
In sum, although changing digitally-mediated network governance may have benefits by allowing more control and value in GPNs, these are only captured by some actors, namely larger firms who achieve savings and greater efficiency. There are also indications that beyond economic benefits the digital is having an impact of the division of labour and is negatively restructuring beneficial relationships, value and increase precarity of some actors.

Power
Tea production in Rwanda occurs through networks of actors (tea pluckers, smallholder farmers, tea processors, tea processor owners, co-operative associations, NGOs etc). An exploration from ANT perspectives would also include non-human actants 6 in the actor-network. In Rwanda, we can identify a number of digital technologies as key actants in shaping networks -smartphones, farmer databases, digital weighing machines and mobile money -that had been brought into, and orientated networks.
Following core ANT approaches we discuss how the quite different ambitions and activities of different actors and firms (profit, farmer livelihoods, national development goals, political goals, donor goals) are aligned and how alignment occurs, particularly with respect to the role of the digital.
In interviews, managers of commercial tea processors were often frustrated at the lack of impetus in the sector in transitioning to become commercially orientated. There were many complaints about the inefficiencies of smallholders, co-operatives, government etc. For these commercial managers, the recruitment of digital technology into discourses was central to articulating a vision for the commercial future of the sector, and an important element of mobilisation for driving towards a fully a commercial sector. For example take the words of one general manager discussing future technology developments.
"We are also adding another sort of information components to that system, where a farmer gets his pay slip on his mobile phone in addition to the green leaf weighing slip. In the future we look at setting up some kind of a platform where they can send users a USSD code and communicate with the factory and access the database there, look at their account and things like that.
Here the digital (USSD, mobile phones, databases) are recruited into the network as a tool to envision and describe the future goals and hopes. Not only do these digital actants solidify the visions of the manager, but suggest inevitability, acting as a mobilising and stabilising force for certain visions.
Secondly, where actor-networks were more integrated, we increasingly saw how norms and rules of production become materialised through the act of inscription (Latour 2005). For digital technology, a particularly studied element of inscription has been in how the embedding of processes into systems, software, or algorithms stabilise certain processes or rules (Spinuzzi 2008). Such digital technologies are important when they circulate because they provide a means for these processes and rules to spread.
One example of inscription and circulation in the Rwandan case was how digital field weighing devices had been integrated with information systems The ways that integration was undertaken in Rwanda was that software and technology was transferred from its Kenya, where it had been custom designed and deployed for the local Kenyan context. Much of Kenya's tea sector is defined by the power of a large umbrella institution (called Kenya Tea development Agency -KTDA) which trades, controls and monitors much of the tea production on behalf of smallholders. This is important, because when field weighing integration was transferred from Kenya into Rwanda, those structures that were inscribed in the technology in Kenya became problematic in Rwanda.
In Kenya, the operation of the software was designed for the KTDA to be able to better manage their tea production from the top-down. Transferred to Rwanda affordances of this system favoured large tea firms, allowing them the ability to input and access information also from the top-down. However, in Rwanda, where there are more independent and active co-operatives who interact with private tea factories., the software was not set up for smaller producers and co-operatives to be afforded access.
Given that these systems were already well established, tested and functionally working from the perception of the large tea firms, there has been reluctance to modify the software and its functions.
In sum, in ANT perspectives, the digital is highly entwined with the actors involved in tea. Digital technologies may be recruited into networks to exert control, as was shown in the illustration above with recruitment of digital actants into articulations of technology. Circulations of digital technologies also had the power to change actor-networks, in how norms inscribed in the digital, and can affect the stability of actor-networks. Digitally-driven shifts in governance and value  Critical approach to new value distribution enabled by digital-exclusionary, spillover effects, who are the winners and losers? Power ANT -Non-human actants  Considering a central role of the digital in the constitution of actor-networks ANT -Translation processes  Focus on translation processes to better explore power and circulation The concept of embeddedness is important for GPN analysis in exploring the coupling between durable economic relations in networks and territories. However, as illustrated in the Rwandan case, there are questions around how embeddedness as a concept is best used. With the digital becoming more prevalent, there is the possibility of a range of diverse network relations through which durable interactions occur (particularly as digital technologies make non-territorial networks easier to maintain).

Discussion
As we highlighted previously, notions of network embeddedness were influenced by Castells' 'space of flows' and 'space of places', and our observations are in line with some of the critiques that have been made of Castells' concepts in the last decade "It seems clear today that the radical dichotomy of the space of flows and the space of places suggested in the late 1980s...has dissolved. Now the connections between the two types of space are much more complex and flexible" (Stalder 2006 p.152) In this quote, Stalder highlights that the simplistic categories of 'network' and 'place' may need to be better refined 7 . This critique serves as an insight for GPN analysis of "network" and "territorial" embeddedness which mirror Castells dichotomy. It can be problematic that network embeddedness becomes catch-all for a whole slew of durable relations, which are not necessarily coherent and consistent. Recent work by Coe & Yeung(2015) on the varying forms of GPN aggregation and Gereffi's discussions of importance of contrasting South-South value chains (Gereffi 2014, Gereffi & Lee 2012) also allude to this increasingly problematic reification of network embeddedness.
We also saw how the digital infrastructure itself was a key element in orientating embeddedness.
Economic relations were embedded in the socio-cultural norms, but network embeddedness was also influenced by the forms of digital infrastructure (i.e. digital integration and digital fragmentation in the Rwandan case). Further research drawing on the information infrastructures literature could be valuable to embeddedness -exploring forms of ICT, applications, codes, standards and algorithms etc. -that can themselves be 'invisible mediators of action' (Bowker & Star 2000).
Where the digital has been associated with shifting distributions of value, existing literature has largely taken an optimistic view focussing on the creation and enhancement of economic rent through new efficiencies and services, but without sufficient empirical support. As outlined in the Rwandan case, new digital elements in network governance can lead to new efficiencies but these gains may be captured outside the places where these efficiencies occur. We also highlighted that the digital had knock-on effects of the division of labour, and the way digital information systems has led to diminished responsibilities for tea co-operatives in the Rwandan case. There is a handful of other recent cases that have explored the effects of the digital on production in such settings, and they also highlight similar impacts of the digital (Graham 2014, Murphy & Carmody 2015: first it can reduce flexibility, sidelining important alliances of producers as opposed to removing intermediaries. Second, integration of digital services may be more exclusionary, due to requirement of digital systems. Third, digital resources that solve a problem, are liable to result in gains far away in the value chain. Thus, the key to a more critical approach, is to reflect on how the digital affects wider chain relations, exploring both intended and unintended effects, and with a closer scrutiny of how the digital is reconfiguring division of labour, and where value is actually captured in netowrks. We this need to pose questions of who benefits, and who does not (in practice) from the digital, and how digital mediated exchanges reconfigure power and relationships in GPNs (Graham 2014).
By bringing in non-humans actants more explicitly into GPN analysis of economic relations, we have been able to explore diverse perspectives on power. Current GPN theorisations use ANT to integrate the analytical concept 'actor-network', but this neglects the core of ANT thinking underplaying the ways that ANT theorises power. Our critique here mirrors Law's(1999) disappointment at the way other fields have appropriated ideas from ANT which have folded a rich set of ideas into a single concept, the 'actor-network' "like some kind of monster, the term 'actor-network' grew, and it started, like a theoretical cuckoo, to throw the other terms out of the nest. Which, with the privilege of hindsight, seems both significant and ominous"(ibid. p.5) In our illustrations we have shown how ANT can provide a useful lens for the bringing into being of actors, norms and networks, through close analysis of the processes of translation, that is the ways by which networks of both human and non-humans are recruited, mobilised and cohered in networks (Murdoch 1998). Translation is at the very core of ANT exploration of understanding how some actors are able to exercise power over others (Law 1999). For the digital, exploring translation processes, is a matter of analysing the alliances and agreements that shape certain digital technologies, which ultimately offers potential to understand power and circulation of ideas in networks.

Reconsidering the digital
This work aligns with a growing literature which has explored less visible processes or artefacts involved in production networks (such as quality, standards and logistics) (Coe & Hess 2013, Ouma 2010, Ponte & Gibbon 2005. These elements, due to the fact that they are "far less tangible and visible in their imprints on the networks" (Coe & Yeung 2015 p.54) have tended to have been less studied in GPN work compared with explorations of agency and firms. Clearer theorisations are integral in better integrating these understudied facets of productions networks.
Through a theoretical examination, this paper has explored the role of the digital in the three conceptual categories of GPNs -embeddedness, value and power. Embeddedness explores the underlying durable norms in territories and networks that shape production relations, where the digital has been seen as a background element that enables durable interactions. We have linked value to network distribution of rents or surpluses and explored the more overt mechanisms used to support exchange. Here the digital has been seen in a mainly positive light in supporting networked transformation through processes such as disintermediation. Integrating the digital into the category of power prompted us to look toward actor-network theory, but there has been barely any substantive use of actor-network theory in studying the digital in GPN literature.
Drawing on empirical illustrations, we have highlighted how the digital could better be conceptualised in GPNs. For embeddedness, complicating network embeddedness to incorporate the complexity of digital networks facilitate is important, as is a consideration to the role that the digital itself plays in shaping durable economic relations. For value, the existing literature suggests a tendency to lean towards idealised visions of the digital and its effect on value. A more critical approach to analysing how the digital influences network exchanges could improve understanding of which actors gain (or not) from the digital, and how it impacts the labour process. For actor-network theory, better consideration of non-humans within actor-networks is a vital starting point to better consider the co-constitution of networks and the role of the digital. Essential to analysis is to position the network as a process, and explore the translation processes by which actor-networks come into being. Figure 3 illustrates these recommendations by superimposing our exploration of the digital onto Henderson et al's(2002) original GPN model. We still take the digital to be a cross-cutting element of GPNs, but we then look to explore how the digital impacts upon the three conceptual categories outlined in key GPN work. Empirical work suggests that all three elements overlap to some degree in that they all try to integrate more complex and materially rich perspectives of the digital in GPNs.
Although we have specifically focussed on the broader GPN position which fits out empirical interest, our conclusions can also be useful for the wider family of models in economic geography that adopt relational frameworks (such as GVC/GCC research). In sum, we see this reconsideration of the digital in GPNs as a starting point. It allows us to make richer theorisations of the digital which compliment perspectives of global production at multiple scales and in constant interplay between networks and territories. The directions and empirical approaches outlined offer clearer directions for future studies to think more deeply about the growing role of the digital in economic production both theoretically and empirically.