The Politics of Digital Agricultural Technologies: a Preliminary Review

The Politics of Digital Agricultural Technologies: A Preliminary Review

Sarah Rotz* , Emily Duncan, Matthew Small, Janos Botschner, Rozita Dara, Ian Mosby, Mark Reed and Evan D.G. Fraser

Abstract

Digital technologies are being developed and adopted across the agro-food system, from farm to fork. Within decision-making spaces, however, little attention is being paid to political factors arising from such technological developments. This review draws from critical social sciences to examine emerging technologies and big data systems in agriculture and assesses some key issues arising in the field. We begin with an introduction and review of the so-called ‘digital revolution’ and then briefly outline how political economy is effective for understanding major challenges for governing technologies and data systems in agriculture. These challenges include: (1) data ownership and control, (2) the production of technologies and data development, and (3) data security. We then use literature and examples to consider the extent to which the political and economic landscape can be shifted to support greater equity in agriculture, while reflecting on structural challenges and limits. In doing so, we emphasise that while there are significant systemic tensions between digital ag-tech development and agroecological approaches, we do not see them as mutually exclusive per se. This article intends to provide decision-makers, practitioners and scholars from a wide range of disciplines with a timely assessment of agro-food digitalisation that attends to political economic factors. In doing so, this article contributes to policy and decision-making discussions, which, from our perspective, continue to be rather technocentric in nature while paying little attention to how digital technologies can support agroecological systems specifically.

© 2019 The Authors Sociologia Ruralis published by John Wiley & Sons Ltd on behalf of European Society for Rural Sociology This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made. Sociologia Ruralis, Vol 59, Number 2, April 2019 DOI: 10.1111/soru.12233 204 Rotz et al.

Introduction

nvironmental problems such as climate change and water scarcity are ex- Epected to make it more difficult and expensive to produce food in the future. Farmers, food scholars, and decision makers alike have argued in recent years that sustainably producing enough healthy food will be one of the most pressing challenges of this century (De Schutter 2012; Marsden 2012; Fraser et al. 2016). One proposed means of addressing this challenge is through ‘digital agriculture’ – defined, broadly, as the application of big data and precision technology systems in agriculture.1 By utilising digital agriculture through introducing a suite of automated, data intensive ‘precision’ technologies, some propose that agriculture could produce more food, on less land, and with fewer inputs (Foresight. The Future of Food and Farming 2011; Franks 2014). In part, the digital agricultural revolution is evolving through a combination of technologies that make use of both cloud computing and the Internet of Things (IoT) while drawing on the immense amount of farm data that modern agricultural operations are now generating. For instance, in 2014, an average of 190,000 data points were produced per farm, per day, and by 2050, experts have predicted that each farm will produce around 4.1 million data points daily (Meola 2016). Although these are global averages, and the vast majority of the data would be generated by farms in the industrialised world, the expectation being that farms in the global south will soon be utilising data-rich technology (World Economic Forum 2018). Emerging technologies such as yield monitors and maps, global positioning systems (GPS), remote sensing, variable rate application (VRA), and robotic milking machines each record their own collection of biophysical and production data on the farm, which helps farmers to customise their practices (Wolfert et al. 2017). As well, digital and cloud- based record keeping, and farm management and decision support software are now available for farms of all sizes and production systems, with many farmers af- firming that these technologies save time, money, and improve their quality of life. There is currently an emerging scholarship exploring the ethical and social ram- ifications of disruptive innovation in big data across a range of sectors, but to date, there is a lack of scholarship on the way in which these technologies may affect agriculture. For example, social scientists have explored the consequences of surveillance through big data (Lyon 2014), the use of big data through social media (Schroeder 2014; Brooker, Barnett, and Cribbin 2016; Felt 2016), and other ethical questions around big data (Zwitter 2014; Illiadis and Russo 2016). Again though, work exploring food system applications of big data remains scarce, and the scholarship that does exist has typically focused on the scientific and/or economic dimensions of big data. To a lesser extent, there have been emerging discussions concerning how digitalisation will exacerbate power inequities in the food system (Bronson and Knezevic 2016; Carolan 2016a, 2016b, 2018; Chi et al. 2017; Mooney 2018). Currently though, no comprehensive review of the political economy of agricultural digitalisation has been written. In fact, much of the critical scholarship declares that these technologies are of little benefit to agriculture, and that, instead, we need to look to agroecological solutions to transform the food system. While we

© 2019 The Authors Sociologia Ruralis published by John Wiley & Sons Ltd on behalf of European Society for Rural Sociology Sociologia Ruralis, Vol 59, Number 2, April 2019 Politics of Digital Agricultural Technologies 205 deeply support the values, principals, and practices of agroecology and are concerned about the trend toward corporate concentration that digitalisation facilitates, we – perhaps optimistically – hope that agroecology does not need to operate outside of digitalisation per se. That said, discussions around ag-tech advancement and uptake continue to be highly technocentric in focus, suppressing opportunities for digital technologies to support small-scale, agroecological approaches. Within this context, this review article considers whether, to what degree, and under what conditions big data technologies could support not only large-scale farmers, but agroecological and small-scale farmers as well. Hence, this article was motivated by the authors desire to explore the key power dynamics that are shaping this emerging field. In doing so we argue that while significant tensions surely exist between digital ag-tech and agroecological approaches, they are not necessarily mutually exclusive. This article integrates a political economy lens to examine emerging big data and technology systems in farming. We integrate political economy by making connec- tions between how different technologies, systems, and components have evolved, how evolution reflects the application of these technologies and, thus, we explore which actors are becoming (dis)empowered through their adoption (Friedmann 1993; Fine, Goodman, and Redclift 1994; Clapp, Newell, and Brent 2017): in other words the goal of this article is to assess who is currently winning and who is losing through the development and application of digital agricultural technologies, and – given this political landscape – consider the extent to which ag-tech can also support agroecological approaches. First, we define this political economy lens and outline the observed trends of digitalisation in agriculture; second, we review the political economic dimensions surrounding three main issues emerging in digital agriculture; third, we reflect on the extent to which digital agricultural technologies can evolve more equitably under the political economic conditions assessed in section two. These three main areas of concern for the field include: (1) data ownership and control, (2) the production of technologies & data development, and (3) data (cyber) security. In doing so, we embed discussions of ‘the digital agricultural revolution’ into a wider context that acknowl- edges the politics of data development (Kitchin 2014; Bronson and Knezevic 2016). Because these issues have been underexamined in scholarship and decision-making thus far, our examination of the political economy of big data, agricultural technology, and smart farming provides a unique and much-needed contribution to existing discussions linking food and big data studies.

Review methodology

This review article is based on the results of an interdisciplinary scholars’ workshop followed by a systematic review of the scholarly literature in agro-food and big data. The workshop included scholars from disciplines such as geography, computer science, history, community and criminal justice, and food studies. First, the research team conducted a workshop to explore the gaps in current analyses. We then created a bibliography of scientific papers using the following key words connected to agriculture: big data, digital(isation), precision farming, cybersecurity, technology, and

© 2019 The Authors Sociologia Ruralis published by John Wiley & Sons Ltd on behalf of European Society for Rural Sociology Sociologia Ruralis, Vol 59, Number 2, April 2019 206 Rotz et al. smart farming. The vast majority of the literature was published between 2010 and 2018. Due to the preliminary nature of the review, white papers and select media articles were also included. Senior scholars on the research team then added supple- mentary readings and provided summations based on their disciplinary background. After this initial phase of review, we sorted the readings into a few broad fields: agriculture 4.0, data governance, and precision agriculture. The field of agriculture 4.0 helped to define digital agriculture broadly, while the other more specific fields divided the literature into dealing with challenges of data itself or precision agriculture referring to the tools used to collect data. After reflecting on key themes that arose through both the workshop and the readings, themes that were cross-cutting to these three fields were selected. The themes were then circulated to the team for feedback and revision, which make up the three main areas of concern laid out in the article: (1) data ownership and control, (2) the production of technologies and data development, and (3) data (cyber) security. Throughout this process, consideration of the political economic dimensions was found to be an important gap in the field.

Digital agriculture: what is it?

Before elucidating the framework for this review, it is necessary to briefly outline the suite of digital technologies that comprise the field of digital agriculture. Ideally, sensors that collect on-farm data in near real time should be integrated with ‘in- telligent’ farm machines, such as smart tractors and automatic milkers that utilise sophisticated algorithms to allow farmers the ability to be more precise with their application of inputs and more knowledgeable about agroecological conditions (Wolfert et al. 2017). For instance, the ‘smart tractor’ may actually be a networked ‘swarm’ of autonomous mini-tractors that use GPS and sensor technology to determine where they are in the field and then ‘learn’ to plant the best suited seed for that particular agroecological context and supply it with the most efficient amount of fertiliser – thus reducing pollution caused by excessive fertiliser application and crop loss caused by under application (Gebbers and Adamchuk 2010). Similarly, automatic milking facilities are becoming common on many farms across North America and Europe (Hansen 2015; Schrijver 2016; Shortall et al. 2016). This is particularly true in Canada where, under supply management, dairy farmers receive a fair price for their product and have greater price stability overall, allowing them to better manage rises in capital costs (Hansen 2015; Schewe and Stuart 2015). Generally speaking, automatic milkers monitor the health and wellbeing of dairy cows in near real-time. They identify ailments and diseases proactively, such as mastitis, and help maintain both productivity and herd health while reducing inputs such as feed and an- tibiotics (Eastwood et al. 2012). Several scholars suggest that the result has been improved cow health and high-quality products with a smaller environmental foot- print (Eastwood et al. 2012; Shortall et al. 2016). Other sectors have similarly seen a growing uptake of these precision technologies. In the Netherlands, for instance, 2015 saw precision techniques being used to manage ‘65 per cent of the nation’s ar- able farmland, up from 15 per cent in 2007’ (Carolan 2017). Big data is increasingly being used to monitor animal health, supporting early detection of animal disease,

© 2019 The Authors Sociologia Ruralis published by John Wiley & Sons Ltd on behalf of European Society for Rural Sociology Sociologia Ruralis, Vol 59, Number 2, April 2019 Politics of Digital Agricultural Technologies 207 and preventing or minimising adverse health impacts. At the herd scale, sensors and other technologies (including microfluidics, sound analysers, image-detection techniques, sweat and salivary sensing and serodiagnosis) are now being used to monitor risk factors that could identify anomalous environmental conditions, physiological parameters and animal behaviours that could lead to early intervention to prevent or detect diseases. For example, sound analysis is being used to identify respiratory disease in pigs in Europe (Ferrari et al. 2008) and detect stress in laying hens in South Korea (Lee et al. 2015), and biosensors are being used for early detection of respiratory disease pigs in the UK (Cowton et al. 2018) and in calves in Japan (Nogami et al. 2014). Big data is also being used at the level of veterinary epidemiology, iden- tifying high risk populations so that surveillance and monitoring can be targeted efficiently (Van der Waal et al. 2017). Many countries now have mandatory animal traceability programmes to track farm animal movements e.g. the European Union’s Trade Control and Expert System and the United Kingdom’s Cattle Tracing System (Green et al. 2018; Bajardi et al. 2012). These systems generate significant amounts of data that are increasingly being analysed and modelled to provide early warning and rapid response in the event of an outbreak e.g. epidemiological modelling using agent-based models of Foot and Mouth Disease transmission in Australia (Bradhurst et al. 2015) and network analysis of livestock movements in France (Lal Dutta et al. 2014). In Switzerland, genetic and production data are routinely linked to identity and movement databases to identify clusters of on-farm deaths and stillbirths that may indicate the emergence or re-emergence of disease (Struchen et al. 2015), and similar moves are being considered in Ireland (Barrett 2017). Of course, these capital-intensive technologies also promise to be lucrative and this has captured the attention of companies, policymakers, and investors alike. For instance, in Canada the Finance Minister’s ‘Advisory Council for Economic Growth’ published a major report in 2017 on how to grow Canadian middle-class incomes. One of their key recommendations was to invest in agro-food sector innovation. Similarly, the UK Government’s £4.7 billion ‘Industrial Strategy Challenge Fund’ includes artificial intelligence and data as one of four challenge areas, with specific schemes focusing on precision agriculture. As well, venture market investment in agricultural technologies has increased 80 per cent annually since 2012 (Sparapani 2017). Similar investments are being made in the United States and across the EU. Future projections predict that the precision agriculture technologies market will exceed $10 billion globally by 2025 (PR Newswire 2017). While these technologies are eagerly promoted by companies, their benefits for farmers themselves are, in many cases, less clear. In many cases, farmers are scep- tical of the benefits and weary of investing in an expensive set of technologies of questionable value. Within nearly all agro-food sectors, farmers have effectively no control over the prices they receive for their crops, which has made the affordabil- ity of these technologies a major concern (Schewe and Stuart 2015; Rotz 2017). On the issue of exclusion, the growing gap between the technological ‘haves’ and ‘have nots’ may only exacerbate the longstanding progression of economic polarisation between small- and large-scale farmers, which has already led to rapid declines in medium-sized farms (Kirschenmann et al. 2004). Meanwhile, even for those who have

© 2019 The Authors Sociologia Ruralis published by John Wiley & Sons Ltd on behalf of European Society for Rural Sociology Sociologia Ruralis, Vol 59, Number 2, April 2019 208 Rotz et al. managed to adopt such technologies, significant barriers remain around data access, management, and analysis.

Digital agriculture: Historic & political dimensions

Political economy has been a common conceptual theory to understand changes in agriculture as it evaluates the influence of economics on social and political systems. In the past, political economy has been used to investigate the transition from ‘local farming practices to concentrated corporate-capitalist agricultural production’ (Friedland et al. 1991, p. 3). Using this lens, we can begin to unpack the politics and power dynamics that lie behind the use of digital agricultural technologies. Political economic theory attempts to address the challenges in understanding relationships of power between agriculturalists, agribusinesses, the state, and non-agricultural institutions (Friedland et al. 1991, p. 26). This form of analysis presents an opportunity to reflect on the motivations of different actors that have contributed to the development of digital agriculture, and, subsequently, how researchers and policymakers might proceed. By interrogating some of the most pressing, and arguably most political, issues affecting the adoption of digital agricultural technologies by farmers (including ownership, technology development, and data/cyber security), we can better understand the trajectory of digital agriculture and its effect on power relations between a diverse range of actors across the agro-food system. In certain respects, the digitalisation of farming is the most significant change to occur in the food system since the Green Revolution of the 1950 s and 1960 s (Friedmann 1993; Moseley 2017). It is widely understood that the growth in corporate power in agriculture has driven up on-farm costs (Qualman 2001; Clapp and Fuchs 2009). Meanwhile, commodity prices have been both volatile and relatively stagnant (on average), forcing many farmers in North America (Basok 2002; Smithers et al. 2005; Skogstad 2007) and Europe (Flaten 2017; Neuenfeldt et al. 2018) to expand production or exit the industry. In turn, across North America average farm size has grown by 65 per cent between 1940 and 2016 while the total number of farms has fallen by 66 per cent over this same period (Statistics Canada 2016; USDA 2016), with similar trends occurring across Europe (Eastwood et al. 2010; van der Ploeg et al. 2015). As the average age of farmers reaches 57-years-old in North America and approximately 55 in Europe, a major concern moving forward is how to ensure that the incoming generation is able to practice farming in ways that are both economically viable and ecologically sustainable (European Commission 2015). As farm size continues to grow, a similar concern has arisen for rural communities more broadly. That is: fewer, larger farms have led, in many cases, to a hollowing out of rural communities as both rural populations and the tax base needed to support services in rural communities declines (Bock 2016). Given this background, we ask, do digital agricultural technologies represent a continuation of historical trends? The most obvious trend is that these technologies will bring about greater market integration and corporate concentration that will exacerbate the farmer debt/income crisis and further exclude small, peasant and agroecological farmers from participating in agro-food production. In effect, as political

© 2019 The Authors Sociologia Ruralis published by John Wiley & Sons Ltd on behalf of European Society for Rural Sociology Sociologia Ruralis, Vol 59, Number 2, April 2019 Politics of Digital Agricultural Technologies 209 economy is a lens through which to understand the relationship between power and technology, political economy is important for understanding the ongoing process of ‘elite capture’, wherein the benefits of new technologies are more easily captured by powerful actors (Hornborg 2001). Of course, while market integration, exclusion, concentration, and elite capture are not new phenomena, it is essential to consider how the economic characteristics of new digital technologies may impact these dynamics across different agro-food sectors. For instance, where technologies are expensive (e.g., $500,000-1 million for a precision tractor) and crop prices are very low and unstable, it is extremely risky for farmers to adopt new technologies, especially in ways that are economically sustainable. In contrast, the Canadian dairy sector is ‘supply managed’, which is a system that helps to stabilise prices for farmers by setting production quotas and minimum price guarantees. This allows farmers a more stable and predictable income within a highly unstable industry. So, while technologies are still capital intensive in dairy (such as automatic milking systems), farmers are better able to manage their on-farm costs according to their annual farm-gate revenues. As well, supply management allows Canadian farmers to maintain smaller herd sizes than in the USA, which is more economically amenable to robot adoption (Rotz et al. 2003). For instance, the average herd size in Ontario is around 70 cows which is roughly what a robot can handle, whereas, the average herd size in the USA is 223 (Statistics Canada 2016; USDA 2016). Herd sizes vary significantly across Europe (from 15 in Poland and Austria to 160 in the Czech Republic), but generally, the uptake of robotic milkers on European dairy farms developed much earlier than in the USA and Canada. At the same time, while farmers might be motivated to adopt new digital technologies on the promise that digital solutions will make them more profitable (alongside the labour-saving benefits of digital agriculture), there are other political and economic drivers that make the adoption of digital technologies more complex. Not only is cost a concern, but farm debt is equally pertinent. Across North America and Europe, farm debt continues to climb while overall profitability declines (Gloy and Widmar 2014; Holtslander 2015; van der Ploeg et al. 2015). As well, real and per- ceived applicability is a significant barrier to technological adoption, which, together, are exacerbating inequities for smaller scale farmers. For instance, farmers have de- scribed first-hand how asset constraints produce inequities in accessing available infrastructure, equipment, resources, and software, which has impacted their ability to participate in digital agriculture altogether (Rotz 2017). In the following section we explore these issues in more detail, with a focus on three key concerns; data ownership and control, the top-down nature of agricultural technology production, and data security.

Key issues in digital agriculture

As several food scholars have shown (Lang 2003; Clapp and Fuchs 2009), political and economic forces of industrialisation have had a major influence on global agriculture. For instance, currently, the Big Six: Bayer, Monsanto, Dow, Dupont, ChemChina, and Syngenta (which are merging into the Big Three, with the recently

© 2019 The Authors Sociologia Ruralis published by John Wiley & Sons Ltd on behalf of European Society for Rural Sociology Sociologia Ruralis, Vol 59, Number 2, April 2019 210 Rotz et al. completed Bayer-Monsanto merger) have captured over 75 percent of the global agricultural input market (Fraser et al. 2016; Clapp 2017). While the food sovereignty movement has risen in resistance to these trends (Desmarais and Wittman 2014), food and farming systems continue to be dominated by large-scale commodity production. Notably, these forces of industrialisation have been central to the development and expansion of digital agriculture (Mooney 2018). One clear implication of the concentration of corporate power is that the vast majority of digital technologies developed thus far are aimed at the needs of large-scale, capital rich farmers, which has direct and indirect impacts for small and mid-sized farmers. Meanwhile, corporate actors are looking to re-shuffle the agro-food landscape in their interest, which will impact farmers of all sizes, as well as farm labourers and food workers. There are three main dimensions through which these political economic dynamics are emerging in the context of digital agriculture; (1) data ownership and control; (2) the production of technology and data development; (3) and data/cyber security. By addressing these dimensions, we can then consider whether a more equitable and sustainable development of these technologies is possible for a more diverse range of farmers and food providers.

Data ownership and control

At the core of the digital agricultural revolution is the use and collection of data to drive management decisions. In this way, data in agriculture is no different than data from other sectors insofar as it is collected from multiple sources (often involv- ing a variety of agencies and private sector entities), exists in many different forms, and may be held and processed by different parties. Within this array of stakeholders, data, and data collection and processing user agreements, there are significant differences in the levels of empowerment among the players. A well-known example of this is in the dairy sector where Lely and De Laval milking systems operate in almost exactly the same way, yet data from these two systems are inaccessible and thus cannot be pooled. Broadly, these are referred to as challenges of data interoperability. Although many of the challenges with interoperability are of a practical nature, such as database structure and organisation, other challenges are rooted in larger political issues of data control and management. For example, programming a combine harvester requires significant data configuration on the part of the farmer, the data from which cannot be exported to another firm’s hardware/software. If the farmer does attempt to adopt technologies – and accept the terms and conditions of using these platforms – by multiple firms, the useful data output required for the next piece of machinery would need to be reproduced from scratch. In this sense, the persistence of multiple, private, competing, and proprietary operating systems and tools are exacerbating interoperability challenges for farmers on the ground. As well, legalistic and technical jargon make it difficult for less resourced stakeholders to remain informed about data frameworks and agreements. Meanwhile, laws and regulations, national standards, guidelines, and infrastructure – alongside powerful lobbying tactics (see for instance the corporate lobby against right-to-repair legislation by John Deere, Apple, and Verizon, which is

© 2019 The Authors Sociologia Ruralis published by John Wiley & Sons Ltd on behalf of European Society for Rural Sociology Sociologia Ruralis, Vol 59, Number 2, April 2019 Politics of Digital Agricultural Technologies 211 rooted in data access constraints and data security, such as the risk of a data breach during the repair process) – influence how and to what degree interoperability can be achieved in the interest of consumers, farmers and less powerful firms and organi- sations across the food system. In turn, new regulatory frameworks and legislation – such as right-to-repair – that prioritise consumer rights and capabilities to access and use software and data are crucial. Although interoperability is an ongoing concern for users, it is increasingly being eclipsed by issues of data siloing and vertical integration, which is ultimately rooted in the longstanding tendency toward corporate monopolisation – only now this is occurring with data itself (Lanier 2014). In effect, companies are incentivised to gain control over data and the results of its analysis – specifically by controlling access to information by a) the same people whom are creating the raw data and b) the sup- posed beneficiaries of the proprietary end products. One means of doing so is by having users trade in data ownership for platform access (with cogent examples being Spotify, Mechanical Turk, or Farmers Edge). Many of these platforms are cloud- based and, therefore, use the Internet to store, manage, and process data. And while cloud computing offers ease of access and streamlined data exchange, it exacerbates corporate accumulation of data while transferring ownership from users to service providers (Merritt 2013). As well, it can work to offset certain dimensions of risk onto service users. For instance, in some cases, the individual users (e.g., farmers) of the companies that create the platforms will become more significantly affected by cybersecurity issues than the company itself, allowing companies to remain financially resilient in the face of operational and economic disruptions. Under the structure of cloud computing, data is thus able to be aggregated and centralised in the interests of service providers who are using vertical integration to concurrently hoard data and acquire the start-ups that are creating both agricultural technologies (AI, sensors, drones, robotics, and biotech) along with the analytics, software, and platforms that are enabling data control. Transferring control and ownership away from users also erodes software reliability and availability: examples abound across sectors wherein users are forced to pay increasing rental fees to access the software that their data is held on (Merritt 2013). As data becomes increasingly valuable, companies move to gain control of data generated across the food system, leading to further corporate concentration. Overall, issues of data ownership and control are not just an effect of technology development, but they almost intrinsically prevent interoperability while reinforcing vendor control. While farmers may agree to the terms and conditions of using digital agricultural software, they have little agency in determining consent rights to their data. Digital agricultural companies often assert that ‘farmers own their data’, however, there is no clear definition of what that ownership means, which results in an ongoing lack of informed consent (Custers 2016). In turn, the democratisation of data control is essential, which would prioritise ‘the right of society over the interests of shareholders’ (Mooney 2018, p. 38). Such steps would create opportunities for agricultural production data to serve the social and ecological interests of farmers and consumers. Ideally, such steps would support the rights of farmers to share on-farm data (including microclimate, soil health, and seed data) without fear of data capture

© 2019 The Authors Sociologia Ruralis published by John Wiley & Sons Ltd on behalf of European Society for Rural Sociology Sociologia Ruralis, Vol 59, Number 2, April 2019 212 Rotz et al. by corporations, which is the kind of data exchange that many organic and agroecological farmers are seeking. Meanwhile, issues of data ownership and control are directly linked to how – and by whom – digital agricultural technologies are produced.

The production of technologies and data development

Digital agricultural technology is posing the greatest challenge for farmers in areas where they have little input into and control over the development process – processes that are typically directed by corporate interests. We characterise this as ‘top-down’ technological development, as opposed to ‘bottom-up’ or ‘farmer-driven’ development. In the former case, technology is developed, and data is collected outside of farmer networks and are thus less likely to be directed by the needs of farmers. As a result, many – especially agroecological – farmers argue that technologies currently promoted to them are not effectively or accurately solving the on-farm problems that they are facing. For instance, many farmers report having yield monitors for over ten years without operating them, primarily because they do not have the supports to calibrate the tractor, or they do not have the tools to transform that data into usable decision-making information (Duncan 2018). This is largely the result of how technologies are developed, and for whose interests. Technologies produced with profit as the central driver are often expensive, inaccessible, and effectively unusable for farmers (Lindblom et al. 2017). Increasingly, decision support systems that process on-farm data will not accept data from multiple sources, which locks farmers into a specific brand and system. This is a well- documented issue with products sold by John Deere, a company that at least one published study concludes denies farmer access to critical software and to protocols for maintaining and modifying their equipment (Motherboard 2018) – although a recent copyright ruling has declared new freedoms for vehicle software repair and modification in the US (U.S. Copyright Office 2018). However, the solution here is not straightforward, as the concern is rooted in complex dimensions of corporate power and concentration. Consider open source software: companies have effectively integrated open source software and platforms into their business models (to accumulate data, extract talent and pressure software developers) in order to create new – often proprietary – innovations. Take, for example, TensorFlow, a software library developed internally by Google, then released open source in 2015. TensorFlow is now being used by the vast majority of AI start-ups for machine learning applications. The research and development outputs of TensorFlow – which are highly valuable for Google – are largely a product of its accessibility and widespread use. Indeed, its ‘learnings’, and hence its value, are derived through it’s open source system. In this way, open source technologies simply do not offer a serious challenge to the status quo in the absence of the kinds of structural shifts necessary to regulate corporate integration. This is why Richard Stallman (2018) was so steadfast in his argu- ment that open source, unlike ‘free’ software, lacks a political imperative – similar to arguments made in the data justice community (Dencik et al. 2016; Taylor 2017). At the same time, top-down technologies are often highly specific. Many proprietary farm management systems focus on very specific production systems, such as

© 2019 The Authors Sociologia Ruralis published by John Wiley & Sons Ltd on behalf of European Society for Rural Sociology Sociologia Ruralis, Vol 59, Number 2, April 2019 Politics of Digital Agricultural Technologies 213 a standard method of soybean cultivation, a vineyard management system, or a single row crop or a veggie truck farming system, which is not conducive to biodynamic and agroecological production systems (Cosgrove 2017). Furthermore, each technology tends to solve particular issues and activities, making the bundle of technologies extremely costly and thus inaccessible to smaller-scale, less capital rich farmers. When farmers decide to rely on a corporate third party to manage data and supply technologies (such as Farmers Edge, AGDATA, and Climate FieldView, to name a few), they should consider the extent to which that party will extract profits from the process–profits that often accumulate at the expense of the farmer. As Carolan (2017, p. 4) highlights, ‘individuals might have access to a tractor’s engine control unit (tECN). Yet this right, in practice, does nothing to reduce farmers’ dependence on agro-food firms, due, in part, to intellectual property regimes’. As a result, farmers report that they are doubtful about whether much of the top- down data and technology is as relevant and effective as their own on-farm knowledge. Early research by Tsouvalis, Seymour, and Watkins (2000) demonstrates that – in the case of yield mapping – farmers who used this technology did not find that they learned anything new from the yield maps they created. The debate continues in the farming community about the usability and utility of top-down software tools, and their capacity to effectively transform on-farm raw yield monitor data into decision making tools that are relevant and applicable to each unique farm context and the specific informational needs of that farmer. Furthermore, farmers have serious concerns about the level of transparency in how calculations are made within the software, leading many to question whether they can trust the outputs. For example, a farmer’s experiential knowledge about the historical conditions and challenges of their soil or livestock cannot be included in the algorithmic calculations made by software. Without being able to integrate such tacit, contextually specific information, many farmers may struggle to trust or see value in the outputs from digital analytical tools (Duncan 2018). Farmers who use agroecological practices and principles are likely to fall into this camp, as agroecology requires in-depth, micro-climate, and often intergenerational knowledge of the agroecosystem (Francis et al. 2003). In turn, data platforms are likely to be developed with data accumulated from large, industrial-scale systems, which further conflicts with the data needs of agroecological producers. A representative of Climate FieldView acknowledged this in explaining how the platform was introduced in Canada, ‘we really did start in a part of Canada that we knew would benefit and work well with the Climate FieldView platform’, which, she goes on to explain to be larger farmers growing grain corn and soybeans (RealAgriculture 2017). Across the supply chain, farmers bear the greatest cost and risk of adopting data gathering technologies – as one poor investment can easily cost them their liveli- hood. It follows, then, that they ought to receive the greatest share of the benefits of the data. Within the current political economic landscape, however, farmers – and agroecological farmers specifically – receive few of the benefits of ag-tech development. Instead, these benefits are being accumulated by developers and service providers, not only through direct profit accumulation from equipment sales and subscription fees, but through the reinvestment and manipulation of the data they

© 2019 The Authors Sociologia Ruralis published by John Wiley & Sons Ltd on behalf of European Society for Rural Sociology Sociologia Ruralis, Vol 59, Number 2, April 2019 214 Rotz et al. have accumulated from both the equipment and from farmers themselves. As these ag-tech companies (who are often also input companies and sell advisory services) receive more agricultural data, they are better able to create profitable products and services, which they then prescribe to farmers – rather than develop alongside farmers. Cooperative and open source models of digital ag development may help to ensure that technology and data are owned directly by the farmer, which can arm them with greater financial power and autonomy. Within bottom-up and/or cooperative processes, solutions are more likely to be directed by farmers, for farmers. The technology and data are thus typically owned and controlled by a community of farmers, sometimes in conjunction with a public agency or university. The focus here is on innovation, which can but does not need to be – defined by capital-intensive forms of technology. Co-operative ag-tech has important implications for small farmers specifically. A recent report by the Joint Research Centre of the European Commission found that, ‘where the field size is small, or when the farmer does not own the technology, specialist contractors, sharing of farming methods and cooperative approaches may be suitable for the use of equipment among different farmers’ (2014, p. 22). This discussion also speaks to the issue of access. As Carolan (2017) details, we must be cautious about our approach to access, as the ability to access something is not the same as having the ‘capabilities to do so in ways that generate material wellbeing’ (p. 17). Hence, achieving access alone does not imply that everyone will have equal access to the supports and resources required to make use of that data. In the case of open data, for instance, a primary concern is that simply opening up data sources without applying checks, balances, and peripheral resources has the ‘potential to exacerbate as much as alleviate injustices’ (Johnson 2014, 263). In the words of Carolan: ‘free access isn’t necessarily fair access’ (2017, p. 20).

Data security

As North Americans woke on Friday, 12 May 2017, a major international news story was breaking in Europe. Parts of the United Kingdom’s National Health Service computer system were down. Companies including FedEx and Deutsche Bank struggled to turn on their computers. Company employees arrived at their terminals to dis- cover nothing more than a red screen and the message, ‘oops, your files have been encrypted!’. What spread through the world’s computer was a piece of ransomware called WannaCry. WannaCry made international headlines by illustrating that cyber security systems were relatively weak given the importance of the economic and social services they support. As technological innovation rapidly evolves, a growing challenge is arising around data privacy and security (Lesser 2014; Sonka 2014). While cyber security has been the object of attention and concern for the past several years, it has tended to focus on threats to privacy, proprietary data, and industrial control systems associated largely with financial services, health services, and various critical infrastructures. Scant attention has been paid to digital agriculture and how risk management and stake- holder engagement should be conceptualised in this sector. As billions of dollars are

© 2019 The Authors Sociologia Ruralis published by John Wiley & Sons Ltd on behalf of European Society for Rural Sociology Sociologia Ruralis, Vol 59, Number 2, April 2019 Politics of Digital Agricultural Technologies 215 accounted for online using digital cryptocurrency like Bitcoin, as the myriad food distribution networks are scheduled, and as our food system is increasingly global- ised while less food is being stored within communities, some argue that food systems are, in turn, becoming more vulnerable to both accidental and malicious attack (Rotz and Fraser 2015; Bogaardt et al. 2016). In some ways, interoperability and cyber security are in tension with each other. Greater interoperability can mean an expansion of threat or attack surfaces, tied to corresponding decreases in the number of segmented or logically closed areas of networks, and increases in the range and numbers of endpoints, gateways, networks, and cloud environments (see Macaulay 2016). While these concerns are understandable, it is also crucial to examine the political and economic utility of such fears. For years, technology manufacturers and service providers have promoted worries over consumer safety and cyber security in order to retain control over repair, recycling, and exchange markets. For instance, in response to the 2017 ‘right to repair’ legislation being proposed in Nebraska, Apple attempted to argue (unsuccessfully) that ‘unauthorised repair will turn the state into a ‘mecca’ for hackers’ (Koebler 2017). In regard to digital agriculture, farmers are having to hack into their own tractors and machinery as, in many cases, they do not have the ‘right to repair’ (Wiens 2015; FarmHack 2018). How, then, is the term ‘hacker’ being applied here? In agriculture, proprietary walls and defences make it virtually impossible for farmers, or even computer pro- grammers for that matter, to make minor repairs – repairs that may end up shutting down high-tech equipment completely. And since it often takes equipment dealers multiple days to order the part and make it out to the farm, urgent on-farm tasks may go unattended to. As a result, farmers themselves now make up the vast majority of the ag-tech ‘hacking’ community: the objective being to simply access the computer system that allows them to make basic repairs, modifications, or collect important information about their crops. After all, time, weather, and cost constraints are some of the reasons why farmers have been repairing their own equipment for as long as there have been farms. As a result, demand for hi-tech tractors are falling as many farmers look to older equipment that does not require computer devices and software to repair (Wiens 2015). So, while data security has been raised as a concern by service providers during de- bates over data control, cyber security more broadly has not yet become a widespread and significant problem in the sector. Nonetheless, as data flows increase – including crop data, livestock health data, and even farmers’ personal online profiles – they may make available the intimacies of farmers’ operations to interested actors down to the square inch of a farm field. The nature and extent of these issues, again, will vary depending on how, and by whom the data is managed and held. A recent deal between farm management software platform, Farmers Edge and PartnerRe (a global reinsurer) illustrates the possible implications. As Farmers Edge CEO Wade Barnes explains: ‘The insights that we get from our data are helpful for growers to make decisions, but there’s a really important opportunity here. When the farmer gets more yield and lowers his costs – that risk management directly flows into the vertical around insurance’ (Cosgrove 2018). While this can support farmers in accessing insurance, we can also envision ways in which this data could be used

© 2019 The Authors Sociologia Ruralis published by John Wiley & Sons Ltd on behalf of European Society for Rural Sociology Sociologia Ruralis, Vol 59, Number 2, April 2019 216 Rotz et al. to deny insurance claims and/or raise policy premiums. This may accelerate the de- cline in family held agricultural operations alongside rising corporate control over agricultural production. Farmers therefore need to consider the potential value of the data they are collect- ing on the farm for vendors of digital agricultural equipment and services and reflect on the specific security and access implications that come along with different kinds of platforms, whether they are proprietary or open source. For example, it requires several years of historical yield data to be able to accurately implement VRA technologies (Zhang 2016). Therefore, if this data was either lost due to a cyber-attack or made inaccessible by the provider, farmers would be in a precarious position, which is especially concerning given the investment required for data gathering technologies. From a risk management perspective, the core business assets of an independent operator may differ considerably from those of a global vendor. While discussing the contours of digital agriculture offer opportunities for shared understanding between individual producers and technology vendors, no clear discussions of the risks related to data collection, data-driven control systems, and data analysis are evolving between users, vendors and decision-makers. Specifically, the comparative nature of these risks, as they reflect different interests and levels of power, have barely been broached. One notable reason for this divide is likely that the entities selling the technology and services are not simply multinational vendors but are increasingly global owners of the new means of agricultural production – data streams and the mechanisms through which these are created and distributed.

Can digital agriculture provide political and economic support for marginalised farmers and food providers?

As scholars have articulated for years, historic and contemporary policies have shaped the political economic conditions through which all technologies emerge, and digital agriculture is no exception. In North America (and in many other contexts) land consolidation, the cost-price squeeze, and pressure towards export-oriented growth have been both a driver and consequence of policy. We acknowledge that these larger trends are likely to continue, which places significant structural constraints around digital agriculture as a tool for small-scale, agro-ecological, and otherwise marginalised farmers and food providers. That said, we suggest that by establishing public support for open source tools alongside greater public funding of technologies, digital agriculture could be made more equitable while supporting greater food system sustainability. Indeed, there are cases where policy has encouraged the development of scale-appropriate technologies and supported access for a diverse range of farmers; for instance, through the enactment of more localised food policies and policy coun- cils that seek to democratise control and prioritise support for community-based food systems. These systems are emerging across North America (e.g. Southern Ontario, British Columbia, and California) and Europe (e.g. the UK, Germany and France). However, there is much more to be done. For example, many local food policies do not yet meaningfully address issues of digitalisation and agro-food technology and consider how small and agroecological farmers could be better supported within this

© 2019 The Authors Sociologia Ruralis published by John Wiley & Sons Ltd on behalf of European Society for Rural Sociology Sociologia Ruralis, Vol 59, Number 2, April 2019 Politics of Digital Agricultural Technologies 217 context, although they could in the future. In the following discussion, we reflect on the extent to which digital agriculture can provide political and economic support to marginalised farmers and food providers, we then consider some specific limitations. First, open source platforms and lower-cost, scale-appropriate technologies are being developed and directed toward small- and medium-sized farmers, which help them gain efficiencies and build on-farm value ‘where the farmer is actually owning more of their own supply chain’ (Cosgrove 2017). One example is the Three Rivers Farmers Alliance, where a network of farmers created a mobile app that allows restaurants, schools, and stores to directly order from local farms who use the software to coordinate harvest, processing, and delivery (Three Rivers Farm 2017). This form of collaborative and democratised software development allows for shorter supply chains and increased communication, which helps to provide greater opportunities for agroecologically diverse and politically marginalised farms to benefit from digitalisation. Again, the presence of open source data and software development alone does not work to shift corporate concentration in agro-tech. In this case, while these networks support local and/or alternative supply chains that operate outside of corporate agro-food assemblages, the political effects are not inherent (Sonnino and Marsden 2006; Knezevic et al. 2013). Open source, collaborative developments such as the Three Rivers Farmers Alliance need to be set within a broader, citizen-led local or regional food policy in order to be sustainable over the long term. A similar trend exists in specialty crop sectors or in areas where lower cost (COG- Pro, AgSquared, or Raspberry Pi automation), co-operative, open source (Farm Hack, FarmOS) and scale-appropriate technologies (Naio Tech’s ‘OZ’ or the Wall-Ye V.I.N. robot) are proving effective – such as viticulture and organics – as well as in sectors looking to reduce labour costs, such as horticulture. Notably, these technologies are ‘creating an alternative for a number of production systems that just simply aren’t represented… [T]he goal isn’t to necessarily replace but make other systems compat- ible with one another’ (Cosgrove 2017). Again however, if public agro-food policy is not developing goals and objectives around how such technologies could be nurtured and supported for local, agroecological farmers, these technologies will remain in the hands of those able to pay, and as these technologies evolve, economic divides will only deepen across the farm-base. Open source technology and data sharing – especially those working outside of the corporate realm of precision agriculture and digitalisation – seem to be some of the more readily effective components of digitalisation for agroecological and small- scale farmers. This is also the case for farmers as a whole. For instance, entities such as Farm Hack, FarmOS, ISOBlue, and AgriLedger – alongside publicly funded platforms – are, in different ways, working to overcome barriers of corporate compe- tition, data siloing and proprietary ownership in the interest of farmers. Farm Hack is ‘a worldwide community of farmers that build and modify’ their own tools and share hacks online and at meet ups (Farm Hack 2018). This model allows farmers to counter elite capture through collectivisation. For instance, Farm Hack’s tool library provides fairly low-cost tech, data software, and innovation solutions for a range of farm sectors, systems and scales. As well, FarmOS software is a free, open source

© 2019 The Authors Sociologia Ruralis published by John Wiley & Sons Ltd on behalf of European Society for Rural Sociology Sociologia Ruralis, Vol 59, Number 2, April 2019 218 Rotz et al. web application open to the whole farm community to support record keeping and farm management. FarmOS is accessible and applicable for farmers of all sizes and sectors (FarmOS 2018). Again, while collectivisation via community and user driven assemblages may help to shift farmer dependence away from the corporate food regime, it will not directly address the broader political trajectory of corporate concentration across the agro-food system, nor within ag-tech specifically. Many scholars and practitioners (Friedland 1991; Friedmann 1993; Clapp and Fuchs 2009; Rotz and Fraser 2015) have argued that addressing these broader issues requires structural shifts in food, environmental and trade policy at the national and global scale. In Canada, for instance, a national food policy is being developed which, some argue, has the potential to significantly increase resources and support for currently marginalised farming systems, research, markets and technologies (e.g., organic/biodynamic, small-scale and new farmers from non-farming backgrounds, perennials and non-commodity crops) (Food Secure Canada 2017; Weiler 2017). However, others are concerned that the policy process will be dominated by corporate power and commodity interests to encourage business-as-usual policy outcomes focused on export production, land consolidation and commodity specialisation (National Farmers Union 2017). It remains to be seen how these tensions will play out in the development of a national Canadian food policy. Yet, it is still important for localised, citizen-led food policies to consider how digitalisation might benefit different kinds of farmers and incorporate mechanisms for locally appropriate technological innovation into those policies. A focus on digital innovation at the local level has the potential to influence broader policy shifts at the national and global scale, particularly in gaining support for digital infrastructure improvements (i.e., broadband internet). Meanwhile, others have called for governance mechanisms such as national man- dates, inter-organisational data charters, social certification schemes, and international treaties in support of open data (de Beer 2016), stressing that open source solutions depend, ‘…in large part on what kind of data is being discussed’ (de Beer 2016, p. 17). Indeed, appropriate governance models for data pertaining to the tra- ditional and culturally specific knowledge of communities ‘is likely to be different from the model for governing qualitative data held by a NGO, and for governing big data collected by a multinational corporation’ (de Beer 2016, p. 17). To be at all effective, governance mechanisms would need to be tied to broader investments in resources and social infrastructure at a local and regional scale ‘that are known to be crucial for innovation, including better education, entrepreneurialism, access to markets, other business and already trained employees’ (Bock 2016). Moreover, governance processes ought to be more closely linked to concepts of justice, rights and forms of knowledge production, links which are being developed through the frameworks of information and data justice – and could thus be more directly tied to food policy development (Johnson 2014; Newman 2015; Taylor 2017). Concerning agricultural data, let us consider who is currently comprising the political space of data production, access and interoperability, outside of agribusiness. Such key players include universities, various farm bureaus, Global Open Data for Agriculture and Nutrition (GODAN), Open Agriculture Data Alliance (OADA) and

AgGateway2 (Porter et al. 2014; Antle, Jones, and Rosenzweig 2017). As well, CGIAR (the Consultative Group for International Agricultural Research), a global partnership engaged in food security research to reduce rural poverty and enhance human health is developing an open-access agricultural data platform (cigar.org 2018). The U.S. Department of Agriculture (USDA) is also currently investing heavily in interoperability, data governance, and open data specifically. What are the potential political implications of these partnerships and investments? Most obviously, even in the public realm, digital ag. will develop through an increasingly universalised lens, which will encourage solutions that are de-contextualised from local economic, social, cultural and ecological contexts. In Canada, for instance, the Office of the Privacy Commissioner has a significant opportunity to partner with Agrifood Canada to invest in public infrastructure for open source technology and data sharing. This could help to ensure that a greater range of farmers have access to – and the capabilities to make use of – emerging technologies in ways that work for their farm. Broadly, these sorts of collaborations have been defined as ‘Agricultural Business Collaboration and Data Exchange Facilities (ABCDEFs)’ (Wolfert et al. 2017). An important question concerning ABCDEFs is, of course whether they ‘will be closed, proprietary systems such as currently Monsanto’s FieldScripts or if these will be more open as proposed by e.g., the OpenATK or the FIspace platform?’ (Wolfert et al. 2017, p. 77). Given the degree of influence that agribusiness, trade agreements, and industry associations have over public agro food governance—or rather, the growing tendency for government to work for the market (Busch 2010) – it is likely that we will continue to see the dominance of proprietary data systems, even as public institutions increasingly involve themselves in data governance. An equally important question here, again, is a question of scale: to what extent can a federally governed project support locally-appropriate digitalisation that is useful for and accessible to agroecological farmers and the agro-food networks they are a part of?

Political economic limitations to enhancing equity in agricultural digitalisation

Although these key public players have been working to develop open standards and protocols while supporting data access and transfer across platforms – as well as building space for community discussion – this does little to shift the political and economic conditions that determine farmers’ dependence on agro-food corporations. Specifically, these processes by themselves are unlikely to empower producers in the face of major corporate interests (Mooney and ETC Group 2015; Bronson and Knezevic 2016; Carolan 2016b). For instance, under the prevailing economic conditions in North America it is easy to envision a future where Apple might own and control nearly all the techno-data software, platforms, and resources available to food producers. This means that, while open source or non-proprietary data and code can help to ‘free’ some farmers to access and use their data in their interests (e.g., to fix or modify their farm equipment, rather than being forced to rely on John Deere techni- cians), they are not necessarily any less dependent on the physical and technological products and services that are embedded in corporate assemblages (Carolan 2016b).

In other words, while open data platforms may give farmers better access and control over the digital aspects of their farm equipment, to generate that data, they still must invest in the physical technologies being developed by an increasingly consolidated and corporately controlled agricultural industry. Moreover, while we agree that open and shared data is an essential means through which we can better guarantee that data remains in the hands of the farm community (de Beer 2016), we must also consider the extent to which open source code and data alone will address the fact that the underlying systems – when created by competing private firms – were never designed for integration to begin with (Wolfert et al. 2017, 77). As mentioned above, the critical technology literature is full of examples of companies that have a long record of engaging in planned obsolescence with previous models of their own products or software, as well as implementing new technologies that lock consumers into a single vendor (LeBel 2016). Until corporate integration and concentration are reined in, perhaps through anti-trust legislation, there is a real concern that these technological trends will result in a continued re- striction of farmers’ options (as opposed to the expansion of options) by agricultural technologies and the firms behind them. Even in the public realm, concerns over scale and locally appropriate technology development loom large. So, while comprehensive public investment and policy into data ownership and governance may be useful, it needs to occur within a locally appropriate, citizen-led policy environment. What is more, though, is that issues of technical interoperability are most urgent for the most isolated rural areas. Currently, many rural areas still lack the basic infrastructure for mobile networks and broadband internet (Janssen et al. 2017). In effect, without significant, decentralised public investment in support for the technical infrastructure needed in rural areas, it is difficult to foresee positive outcomes for the sustainability of these communities as digitalisation evolves. Concerning cyber security, while many are working to better understand data integrity, data security, and cyber threat detection and response (e.g., the human dimensions alongside the operational and sectoral threats and opportunities), this does not seem to be happening within a broader political conversation around data control and ownership (Chi et al. 2017). At the political level, farmers, food workers and consumers ought to be directing the process with government to shape regulatory conditions concerning how and by whom their data is held and controlled. Moreover, ‘innovation’ in food and agriculture has too long been equated strictly with ‘technology’. Of course, on-farm innovation is always occurring and evolving and may include – but is not limited to – the adoption of big data and technology systems. It might also involve the adoption of practices that mimic nature, such as adding natural soil amendments, intercropping, and livestock integration to build soil organic matter and naturally mitigate pests, or the integration of trees and naturalised areas to enhance biodiversity (Gliessman 2007; Pretty 2008). As well, innovative ecologically sustainable and intensive systems are being developed across the globe, such as agro-ecological vertical farming methods (Coleman 2014). In this sense, locally-appropriate policy should support ecologically driven agricultural innovation through both programming and resource provision. Focusing on

© 2019 The Authors Sociologia Ruralis published by John Wiley & Sons Ltd on behalf of European Society for Rural Sociology Sociologia Ruralis, Vol 59, Number 2, April 2019 Politics of Digital Agricultural Technologies 221 community-level solutions should help nurture innovation that is occurring outside of corporate labs and boardrooms.

Data justice for farmers

Data justice has been defined as ‘fairness in the way people are made visible, represented, and treated as a result of their production of digital data’ (Taylor 2017). Through a political-economic analysis of digital agriculture, it is clear that many farmers struggle to see the benefits of the digital data that their farm technologies are producing. Rising corporate concentration has made it such that farmers lack the ability and permissions to access and utilise data while all the power resides with the companies who have the permission and capability to store, manage, and manipulate it (Crawford et al. 2014). Meanwhile, a similar concern raised by Taylor (2017) can be raised in the context of farmers: while many small and lower income commodity farmers are systemically driven to adopt precision and digital technologies, these technologies and systems are often aligned with the models and production systems of large, incorporated, industrial scale farmers, much of which are merely out of reach for smaller farmers. In effect, economies of scale alongside capital accumulation make it possible for larger farmers to access the latest technology, which then reduces their fuel, pesticide and input costs. These increasing technological disparities further economic polarisation and make it increasingly difficult for small and medium sized farmers to participate in agriculture altogether. Meanwhile, low-input, agroecological and specialty producers remain largely absent from the data development sphere of agriculture in North America. This raises questions about which farms and farmers are being (mis)represented as agricultural data accumulates. Finally, while data rights for farmers has been a growing concern, we have shown in this article that bottom-up, locally-appropriate, farmer-driven digitalisation – which reflects the needs of farmers – is just as central to data justice in agriculture. This conclusion supports what others have suggested around the framings of justice more broadly (Taylor 2017). Continuing research that questions the epistemologies of the use of big data (Crawford et al. 2014; Kitchin 2014), particularly focused on agriculture, is needed if we are to better understand ways to incorporate fairness and social justice into digital agriculture. For instance, how might digital technologies impact the most vulnerable in the industry, primarily precarious farm and food workers? As well, big data scholars need to unpack the ‘black box’ that surrounds agricultural data once it leaves the farm. Once these data pathways are better understood, we can start to envision how agricultural data can benefit a diversified agricultural system.

Conclusion

Over the next generation, we face a significant challenge: how to sustainably, safely and nutritiously feed a growing population while addressing major environmental problems, such as climate change and water scarcity. While technology has a role to play, many of the so-called technological solutions are being developed in ways that empower corporate actors rather than supporting independent farmers to make

© 2019 The Authors Sociologia Ruralis published by John Wiley & Sons Ltd on behalf of European Society for Rural Sociology Sociologia Ruralis, Vol 59, Number 2, April 2019 222 Rotz et al. informed decisions about the agroecological system under their management. The current path of agricultural tech may exacerbate inequities for marginalised food system actors, specifically between different sized farmers as well as farmers and agro-food corporations. In this article, we have reviewed the political economy of digital agriculture as it relates to three major challenges in the sector, (1) data ownership and control, (2) the production of technologies & data development, and (3) data (cyber) security. Our discussion of these challenges is not meant to be exhaustive but, rather, to profile and clarify some of the most prevalent political economic concerns arising across the sector in a way that is useful for a wide range of decision- makers and scholars alike. In particular, we note the value of open, co-operative, publicly funded and locally appropriate technology and data systems as first steps in supporting solutions that contribute to data justice for farmers. We emphasise, however, that larger political economic barriers in agriculture surely limit the extent to which digitalisation can support the interests of marginalised farmers and food growers. Much more scholarly and pragmatic work needs to be done, then, if we hope to be better able to understand what data justice means for the agricultural community and how it can be achieved moving forward.

Notes

*Corresponding author. 1. While the literature often conflates terms such as digital agriculture, big data, smart farming, and precision agriculture, the broad term of digital agriculture refers to the use of both big data and precision agriculture and will be the focus of this article. 2. GODAN is a global partnership that promotes the ‘proactive sharing of open data to make information about agriculture and nutrition available, accessible and usable’ (godan.info 2018). The OADA is an alliance that operates with a farmer-focused approach and ‘through a central guiding principle that each farmer owns data generated or entered by the farmer, their employees or by machines performing activities on their farm. They develop open reference systems and protocols for farm communities’ (openag.io 2018). AgGateway ‘is a non-profit consortium of businesses serving the agriculture industry’ with a mission to ‘promote and enable the industry’s transition to digital agriculture, and expand the use of information to maximise efficiency and productivity’ (aggateway.org).

Acknowledgements

This work is supported by the Social Sciences and Humanities Research Council of Canada (SSHRC) and is associated with ‘Food from Thought’, the University of Guelph’s Canada First Research Excellence Fund project.

References

Antle, J.M., J.W. Jones and C. Rosenzweig (2017) Next generation agricultural system models and knowledge products: Synthesis and strategy. Agricultural Systems 155 (June) pp. 179–185. Bajardi, P., A. Barrat, L. Savini and V. Colizza (2012) Optimizing surveillance for livestock disease spreading through animal movements. Journal of the Royal Society Interface 9 (76) pp. 2814–2825.

Barrett, D. (2017) The potential for big data in animal disease surveillance in Ireland. Frontiers in Veterinary Science 4 pp. 150. Basok, Tanya (2002) The farmers’ affliction. Pp. 25–41 Tortillas and tomatoes transmigrant mexican harvesters in Canada (Montreal: McGill-Queen’s University Press). de Beer, Jeremy (2016) Ownership of open data: Governance options for agriculture and nutrition. GODAN summit. (New York City: Global Open Data for Agriculture & Nutrition [GODAN]) Bock, Bettina B. (2016) Rural marginalisation and the role of social innovation; a turn towards nexogenous development and rural reconnection. Sociologia Ruralis 56 (4) pp. 552–573. Bogaardt, M.J., K.J. Poppe, V. Viool and E. van Zuidam (2016) Cybersecurity in the agrifood sector. Capgemini Consulting. Vol. 12. https://www.wur.nl/upload_mm/4/6/a/f74a893e-c829- 4bf3-9884-e357929ff5d6_Cybersecurity in the agrifood sector.pdf. Bradhurst, R.A., S.E. Roche, I.J. East, P. Kwan and M.G. Garner (2015) A hybrid modeling approach to simulating foot-and-mouth disease outbreaks in Australian livestock. Frontiers in Environmental Science 3 pp. 17. Bronson, K. and I. Knezevic (2016) Big data in food and agriculture. Big Data & Society 3 (1) pp. 1–5. Brooker, P., J. Barnett and T. Cribbin (2016) Doing social media analytics. Big Data & Society 3 (2) pp. 1–12. Busch, L. (2010) Can fairy tales come true? The surprising story of neoliberalism and world agriculture. Sociologia Ruralis 50 (4) pp. 331–351. Carolan, M. (2016a) Agro-digital governance and life itself: Food politics at the intersection of code and affect. Sociologia Ruralis 57 (S1) pp. 816–835. Carolan, M. (2016b) Publicising food: Big data, precision techniques of addition. Sociologia Ruralis 57 (2) pp. 1–20. Carolan, M. (2017) ‘Smart’ farming techniques as political ontology: Access, sovereignty and the performance of neoliberal and not-so-neoliberal worlds. Sociologia Ruralis 57 (2) pp. 745–764. Carolan, M. (2018) Big data and food retail: Nudging out citizens by creating dependent consumers. Geoforum 90 pp. 142–150. Chi, H., S. Welch, E. Vasserman and E. Kalaimannan (2017) A framework of cybersecurity approaches in precision agriculture. Proceedings of the 12th international conference on cyber warfare and security 90–95. Clapp, J. (2017) Bigger is not always better: Drivers and implications of the recent agribusiness megamergers (Waterloo: Global Food Politics Group). Clapp, J. and D. Fuchs (2009) Corporate power in global agrifood governance (Cambridge, MA: The MIT Press). Clapp, J., P. Newell and Z.W. Brent (2017) The global political economy of climate change, agriculture and food systems. The Journal of Peasant Studies pp. 1–9. Coleman, R. (2014) Low-Tech Innovations in Vertical Farming: Nairobi, Kenya. Urban Agriculture Magazine, December. no 28 GROW the City. Innovations in Urban Agriculture, pp. 65–67. Cosgrove, E. (2017) How do farm hackers view Venture-backed agtech? AgFunderNews. https:// agfundernews.com/how-do-farm-hackers-view-venture-backed-agtech.html. Cosgrove, E. (2018) Farmers edge partners with global reinsurer to bring farmers custom data-backed insurance products. AgFunderNews. https://agfundernews.com/farmers-edge-partners-global-reinsurer.html. Cowton, J., I. Kyriazakis, T. Plötz and J. Bacardit (2018) A combined deep learning GRU- autoencoder for the early detection of respiratory disease in pigs using multiple environmental sensors. Sensors 18 (8) pp. 2521. Crawford, K., K. Miltner and M.L. Gray (2014) Special section introduction. International Journal of Communication 8 pp. 1663–1672.

Custers, B. (2016) Click here to consent forever: Expiry dates for informed consent. Big Data & Society 3 (1) pp. 1–6. Dencik, L., A. Hintz and J. Cable (2016) Towards data justice? The ambiguity of anti-surveillance resistance in political activism. Big Data & Society 3 (2) pp. 1–12. Desmarais, A.A. and H. Wittman (2014) Farmers, foodies and first nations: Getting to food sovereignty in Canada. The Journal of Peasant Studies 41 pp. 1–23. Duncan, E. (2018) An exploration of how the relationship between farmers and retailers influ- ences precision agriculture adoption. MA Thesis. Department of Geography, University of Guelph. Dutta, B.L., P. Ezanno and E. Vergu (2014) Characteristics of the spatio-temporal network of cattle movements in France over a 5-year period. Preventive Veterinary Medicine 117 (1) pp. 79–94. Eastwood, R., M. Lipton and A. Newell (2010) Farm size. Handbook of Agricultural Economics 4 pp. 3323–3397. Eastwood, C.R., D.F. Chapman and M.S. Paine (2012) Networks of practice for co-construction of agricultural decision support systems: Case studies of precision dairy farms in Australia. Agricultural Systems 108 pp. 10–18. European Commission (2015) EU farms and farmers in 2013: An update. EU Agricultural and Farm Economic Briefs. http://ec.europa.eu/agriculture/rural-area-economics/briefs/ pdf/009_en.pdf. Farm Hack (2018) Farm hack. http://farmhack.org/tools. FarmOS (2018) Introduction - FarmOS.Org. http://farmos.org/guide/. Felt, M. (2016) Social media and the social sciences: How researchers employ big data analytics. Big Data & Society 3 (1) pp. 1–15. Ferrari, S., M. Silva, M. Guarino, J.M. Aerts and D. Berckmans (2008) Cough sound analysis to identify respiratory infection in pigs. Computers and Electronics in Agriculture. 64 (2) pp. 318–325. Fine, B., D. Goodman and M. Redclift (1994) Towards a political economy of food. Review of International Political Economy 1 (3) pp. 547–552. Flaten, O. (2017) Factors affecting exit intentions in Norwegian sheep farms. Small Ruminant Research 150 pp. 1–7. Food Secure Canada (2017) From patchwork to policy coherence: Principles and priorities of Canada’s national food policy. https://foodsecurecanada.org/patchwork-policy- coherence-principles-and-priorities-canadas-national-food. Foresight. The Future of Food and Farming (2011) Final project report: Challenges and choices for global sustainability. The Government Office for Science pp. 1–211. Francis, C., G. Lieblein, S. Gliessman, T.A. Breland, N. Creamer, R. Harwood, L. Salomonsson, et al. (2003) Agroecology: The ecology of food systems. Journal of Sustainable Agriculture 22 (3) pp. 99–118. Franks, J.R. (2014) Sustainable intensification: A UK perspective. Food Policy 47 pp. 71–80. Fraser, E.D.G., A. Legwegoh, K.C. Krishna, M. CoDyre, G. Dias, S. Hazen, R. Johnson, et al. (2016) Biotechnology or organic? Extensive or intensive? Global or local? A critical review of potential pathways to resolve the global food crisis. Trends in Food Science and Technology 48 pp. 78–87. Friedland, W. (1991) Towards a new political economy of agriculture (Westview special studies in agriculture science and policy). (Boulder, CO: Westview Press). Friedmann, Harriet (1993) The political economy of food: A global crisis. New Left Review, 197 pp. 29–57. Gebbers, R. and V.I. Adamchuk (2010) Precision agriculture and food security. Science 327 pp. 828–830.

Gliessman, S. (2007) Agroecology: The ecology of sustainable food systems (Boca Raton: CRC Press). Gloy, B.A. and D.A. Widmar (2014) Long-term trends in farm demographics, debt use, and land ownership: Implications for the financial needs of U.S. farming. A Report for the Farm Credit System Coordinating Committee pp. 1–23. Green, D.M., I.Z. Kiss, A.P. Mitchell and R.R. Kao (2008) Estimates for local and movement-based transmission of bovine tuberculosis in British cattle. Proceedings of the Royal Society B: Biological Sciences 275 (1638) pp. 1001–1005. Hansen, B.G. (2015) Robotic milking-farmer experiences and adoption rate in Jæren, Norway. Journal of Rural Studies 41 pp. 109–117. Holtslander, C. (2015) Losing our grip: 2015 update. National Farmers Union 36 pp. 1–36. Hornborg, A. (2001) The power of the machine: Global inequalities of economy, technology, and environment. Walnut Creek, CA; Oxford: AltaMira Press. Iliadis, A. and F. Russo (2016) Critical data studies: An introduction. Big Data & Society 3 (2) pp. 1–7. Janssen, S.J.C., C.H. Porter, A.D. Moore, I.N. Athanasiadis, I. Foster, J.W. Jones and J.M. Antle (2017) Towards a new generation of agricultural system data, models and knowledge products: Information and communication technology. Agricultural Systems 155 pp. 200–212. Johnson, J.A. (2014) From open data to information justice. Ethics and Information Technology 16 (4) pp. 263–274. Joint Research Centre (JRC) of the European Commission (2014) Precision agriculture: An opportunity for EU farmers-potential support with the CAP 2014–2020. European Union pp. 1–56. Kirschenmann, F., S. Stevenson, F. Buttel, T. Lyson and M. Duffy (2004) Why worry about the agriculture of the middle? Leopold Center Pubs and Papers 143 pp. 1–23. Kitchin, R. (2014) Big data, new epistemologies and paradigm shifts. Big Data & Society 1 (1) pp. 205395171452848. Knezevic, I., K. Landman and A. Blay-palmer (2013) Local food systems - International perspec- tives: A review pp. 1–12 (Guelph, ON, Canada: Ontario Ministry of Agriculture and Food). Koebler, J. (2017) Apple tells lawmaker that right to repair iPhone will turn Nebraska into a ‘Mecca’ for hackers. Motherboard. https://motherboard.vice.com/en_us/article/pgxgpg/apple-tells-lawmaker-that-right-to-repair-iphones-will-turn-nebraska-into-a-mecca-for-hackers. Lang, T. (2003) Food industrialisation and food power: Implications for food governance. Development Policy Review 21 (5–6) pp. 555–568. Lanier, J. (2014) Who owns the future? Simon & Schuster (New York: Simon & Schuster). LeBel, S. (2016) Fast machines, slow violence: ICTs, planned obsolescence, and e-waste. Globalizations 13 (3) pp. 300–309. Lee, J., B. Noh, S. Jang, D. Park, Y. Chung and H.H. Chang (2015) Stress detection and classi- fication of laying hens by sound analysis. Asian-Australasian Journal of Animal Sciences. 28 (4) pp. 592. Lesser, A. (2014) Big Data and Big Agriculture. Gigaom, 11. https://gigaom.com/report/ big-data-and-big-agriculture/. Lindblom, J., C. Lundström, M. Ljung and A. Jonsson (2017) Promoting sustainable intensification in precision agriculture: Review of decision support systems development and strate- gies. Precision Agriculture 18 (3) pp. 309–331. Lyon, D. (2014) Surveillance, snowden, and big data: Capacities, consequences. Critique. Big Data & Society 1 (2) pp. 1–13. Macaulay, T. (2016) RiOT Control: Understanding and managing risks and the Internet of Things. Self-published. ISBN 978-0-12-419971-2. Marsden, T. (2012) Towards a real sustainable agri-food security and food policy: Beyond the ecological fallacies? The Political Quarterly 83 (1) pp. 139–145.

Meola, A. (2016) Why IoT, Big Data & Smart Farming Are the Future of Agriculture. Business Insider. http://www.businessinsider.com/internet-of-things-smart-agriculture-2016-10. Merritt, H.C. (2013) Sharecropping in the Cloud. Jacobin. https://www.jacobinmag.com/2013/11/ sharecropping-in-the-cloud. Mooney, P. (2018) Blocking the Chain. (Berlin, Germany: ETC Group). Mooney, P. and ETC Group (2015) The changing agribusiness climate: Corporate concentration, agricultural inputs, innovation, and climate change. Canadian Food Studies / La Revue Canadienne Des Études Sur l’alimentation 2 (2) pp. 117. Moseley, W.G. (2017) The Limits of the New Green Revolution for Africa: A Political Ecology Critique. Brown Journal of World Affairs XXIII (II). Motherboard. (2018). Tractor Hacking: The Farmers Breaking Big Tech’s Repair Monopoly. YouTube. https://www.youtube.com/watch?time_continue=1&v=F8JCh0owT4w. National Farmers Union (2017) A food policy for Canada. Saskatoon. https://www.canada.ca/en/ campaign/food-policy.html. Neuenfeldt, S., J. Rieger, T. Heckelei, A. Gocht, P. Ciaian and G. Tetteh (2018) A multiplicative competitive interaction model to explain structural change along farm specialisation, size and exit/entry using Norwegian farm census data. In 2018 Conference, July 28-August 2, 2018, Vancouver, British Columbia (No. 277090). International Association of Agricultural Economists. Newman, N. (2015) Data justice: Taking on big data as an economic justice issue. Data Justice, no. March: pp. 1–31. Newswire, P.R. (2017) Precision farming/agriculture market analysis 2014–2025. Research and Markets. http://www.prnewswire.co.uk/news-releases/precision-farmingagriculture-market-analysis-2014-2025–-focus-on-yield-monitoring-field-mapping-crop-641291673.html. Nogami, H., H. Okada, T. Miyamoto, R. Maeda and T. Itoh (2014) Wearable wireless tempera- ture sensor nodes appressed to base of a calf’s tail. Sens. Mater 26 pp. 539–545. van der Ploeg, J.D., J.C. Franco and S.M. Borras (2015) Land concentration and land grabbing in Europe: A preliminary analysis. Canadian Journal of Development Studies 36 (2) pp. 147–162. Porter, C.H., C. Villalobos, D. Holzworth, R. Nelson, J.W. White, I.N. Athanasiadis, S. Janssen, et al. (2014) Harmonization and translation of crop modeling data to ensure interoperability. Environmental Modelling and Software 62 pp. 495–508. Pretty, J. (2008). Agricultural Sustainability: Concepts, Principles and Evidence. Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences 363 (1491): 447–465. Qualman, D. (2001) The Farm Crisis and Corporate Power. (Saskatoon, SK, Canada: Canadian Centre for Policy Alternatives). RealAgriculture (2017) Taking a Closer Look at Climate FieldView. RealAgriculture. https:// www.realagriculture.com/2017/09/taking-a-closer-look-at-climate-fieldview/. Rotz, S. (2017) Drawing lines in the cornfield: An analysis of discourse and identity relations across agri-food networks. Agriculture and Human Values 35 (2) pp. 441–456. Rotz, S. and E.D.G. Fraser (2015) Resilience and the industrial food system: Analysing the impacts of agricultural industrialization on food system vulnerability. Journal of Environmental Studies and Sciences 5 (3) pp. 459–473. Rotz, C.A., C.U. Coiner and K.J. Soder (2003) Automatic milking systems, farm size, and milk production. Journal of Dairy Science 86. Elsevier pp. 4167–4177. Schewe, R.L. and D. Stuart (2015) Diversity in agricultural technology adoption: How are automatic milking systems used and to what end? Agriculture and Human Values 32 pp. 199–213. Schrijver, R. (2016) Precision agriculture and the future of farming in Europe: Scientific foresight study. Scientific Foresight Study IP/G/STOA/FWC/2013-1/Lot 7/SC5 December 2016. pp 1–40. Schroeder, R. (2014) Big data and the brave new world of social media research. Big Data & Society 1 (2) pp. 1–11.

De Schutter, O. (2012) Report of the special rapporteur on the right to food on his mission to Canada. United Nations. Shortall, J., L. Shalloo, C. Foley, R.D. Sleator and B.O. Brien (2016) Investment appraisal of automatic milking and conventional milking technologies in a pasture-based dairy system. Journal of Dairy Science 99 (9) pp. 7700–7713. Skogstad, G. (2007) The two faces of Canadian agriculture in a post-staples economy. Canadian Political Science Review 1 (1) pp. 26–41. Smithers, John, A.E. Joseph and M. Armstrong (2005) Across the divide (?): reconciling farm and town views of agriculture-community linkages. Journal of Rural Studies 21 (3) pp. 281–295. Sonka, S. (2014) Big data and the ag sector: More than lots of numbers. International Food and Agribusiness Management Review 17 (1) pp. 1–20. Sonnino, R. and T. Marsden (2006) Beyond the divide: Rethinking relationships between alternative and conventional food networks in Europe. Journal of Economic Geography 6 (2) pp. 181–199. Sparapani, T. (2017) How big data and tech will improve agriculture, from farm to table. Forbes. https://www.forbes.com/sites/timsparapani/2017/03/23/how-big-data-and-tech- will-improve-agriculture-from-farm-to-table/#8733afc59891. Stallman, R. (2018) Why Open Source Misses the Point of Free Software. GNU. Accessed 13 February. https://www.gnu.org/philosophy/open-source-misses-the-point.html.en. Statistics Canada (2016) 2016 Census of Agriculture. Statistics Canada. http://www.statcan. gc.ca/daily-quotidien/170510/dq170510a-eng.htm. Struchen, R., M. Reist, J. Zinsstag and F. Vial (2015) Investigating the potential of reported cattle mortality data in Switzerland for syndromic surveillance. Preventive Veterinary Medicine 121 (1–2) pp. 1–7. Taylor, L. (2017) What is data justice? The case for connecting digital rights and freedoms globally. Big Data & Society 4 (2) pp. 1–14. Three Rivers Farm (2017) Three river farmers alliance. https://www.threeriverfa.com/. Tsouvalis, J., S. Seymour and C. Watkins (2000) Exploring knowledge-cultures: Precision farming, yield mapping, and the expert - farmer interface. Environment and Planning A 32 (5) pp. 909–924. U.S. Copyright Office (2018) Exemption to prohibition on circumvention of copyright pro- tection systems for access control technologies. U.S. Copyright Office. www.copyright. gov/1201/2018/USCO-letters/. USDA (2016) 2016 Agricultural Statistics Annual. National Agricultural Statistics Service. https://www.nass.usda.gov/Publications/Ag_Statistics/2016/index.php. VanderWaal, K., R.B. Morrison, C. Neuhauser, C. Vilalta and A.M. Perez (2017) Translating big data into smart data for veterinary epidemiology. Frontiers in Veterinary Science 4 pp. 110. Weiler, A. (2017) Health, dignity and human rights: How a national food policy can strengthen justice with migrant farm workers in Canada. https://foodsecurecanada.org/sites/foodsecurecanada.org/files/files/migrant_workers_briefing_note_weiler.pdf. Wiens, K. (2015) New high-tech farm equipment Is a nightmare for farmers. WIRED. https:// www.wired.com/2015/02/new-high-tech-farm-equipment-nightmare-farmers/. Wolfert, S., L. Ge, C. Verdouw and M. Bogaardt (2017) Big data in smart farming – A review. Agricultural Systems 153 pp. 69–80. World Economic Forum (2018) Innovation with a purpose: The role of technology innovation. http://www3.weforum.org/docs/WEF_Innovation_with_a_Purpose_VF-reduced.pdf. Zhang, Q. (2016) Precision agriculture technology for crop farming. Edited by Qin Zhang. (Boca Raton, FL: CRC Press). Zwitter, A. (2014) Big data ethics. Big Data & Society 1 (2) pp. 1–6.

Sarah Rotz * University of Guelph 50 Stone Rd E Guelph Ontario, Canada N1G 2W1 e-mail: [email protected]

Emily Duncan University of Guelph 50 Stone Rd E Guelph Ontario, Canada N1G 2W1

Matthew Small Conestoga College Institute of Technology and Advanced Learning 299 Doon Valley Dr Kitchener Ontario, Canada N2G 4M4

Janos Botschner Conestoga College Institute of Technology and Advanced Learning 299 Doon Valley Dr Kitchener Ontario, Canada N2G 4M4

Rozita Dara University of Guelph 50 Stone Rd E Guelph Ontario, Canada N1G 2W1

Ian Mosby University of Guelph 50 Stone Rd E Guelph Ontario, Canada N1G 2W1 e-mail: [email protected]

Mark Reed Newcastle University King’s Road Newcastle upon Tyne NE1 7RU United Kingdom

Evan D.G. Fraser University of Guelph 50 Stone Rd E Guelph Ontario, Canada N1G 2W1