“If they don’t tell us what they do with it, why would we trust them?” Trust, transparency and benefit-sharing in Smart Farming

Abstract Advances in Smart Farming and Big Data applications have the potential to help agricultural industries meet productivity and sustainability challenges. However, these benefits are unlikely to be realised if the social implications of these technological innovations are not adequately considered by those who promote them. Big Data applications are intrinsically socio-technical; their development and deployment are a product of social interactions between people, institutional and regulatory settings, as well as the technology itself. This paper explores the socio-technical factors and conditions that influence the development of Smart Farming and Big Data applications, using a multi-level perspective on transitions combined with social practice theory. We conducted semi-structured interviews with 26 Australian grain farmers and industry stakeholders to elicit their perspectives on benefits and risks of these changes. The analysis shows that issues related to trust are central concerns for many participants. These include procedural concerns about transparency and distributional concerns about who will benefit from access to and use of “farmers’ data”. These concerns create scepticism about the value of ‘smart’ technologies amongst some industry stakeholders, especially farmers. It also points to a divergence of expectations and norms between actors and institutions at the regime and niche levels in the emerging transition towards Smart Farming. Bridging this divide will require niche level interventions to enhance the agency of farmers and their local networks in these transactions, and, the cooperative design of new institutions at regime level to facilitate the fair and transparent allocation of risk and benefit in farming data information chains.


Introduction
Advances in Smart Farming (also known as digital farming or digital agriculture) and Big Data applications have the potential to deliver a range of benefits, such as improved decision-making, increased efficiency and economic gain and even decreased environmental impact, which could in turn help agricultural industries meet their productivity and sustainability challenges (Sonka, 2016;Everingham et al., 2016;Wolfert et al., 2017).Smart Farming takes advantage of emerging smart machines, sensors and precision agriculture equipment that create vast amounts of real-time farm data (e.g.monitoring animals, soil, water and plants) and uses this data to make more timely or accurate decisions, both on-farm and across the supply chain (Wolfert et al., 2017;Eastwood et al., 2017).Big Data refers to the capability to extract information and insights at a large scale where previously it was economically and technically not possible (Sonka, 2015), through the use of "computerised analytical systems that interrogate extremely large databases of information in order to identify particular trends and correlations" (Keogh and Henry, 2016: 4).Big Data is often described in terms of the 3 Vs: Volume; Velocity; and Variety (Manyika et al., 2011;Kitchin and McArdle, 2016).Big Data applications are already being deployed to improve productivity and profitability in many sectors (van Rijmenam, 2017;Davenport and Dyché, 2013;Kitchin and Data, 2014).Early experiences with Big Data applications in agriculture suggest that their success hinges upon multiple social and technical factors, including the willingness of stakeholders to share and integrate data, enduser acceptance of the technologies, and the existence of protocols for protecting farmers' rights to privacy, data ownership and control (Sonka, 2015;Eastwood and Yule, 2015;Griffith et al., 2013;Kaloxylos et al., 2014;Poppe et al., 2015).Furthermore, Big Data applications have the potential to transform roles and power relationships between stakeholders within the agricultural sector (Wolfert et al., 2017;Bronson and Knezevic, 2016).
Consequently, Big Data applications are socio-technical configurations, their development and deployment being a product of social interactions between people, institutional and regulatory settings, as well as the technology itself (Vines et al., 2013).In this respect, Big Data applications are similar to other agri-environmental decision-support technologies, or farming practices more broadly, that rely upon stakeholder collaboration (Carberry et al., 2002;Jakku and Thorburn, 2010) and trusted local networks and intermediaries that buffer farmers' perceived risks and enhance local benefits (Taylor and Van Grieken, 2015).These dynamics occur within and across nested levels of social, institutional and cultural organisations that are tied to processes of innovation and transitions in society.To address what we argue is presently an empirical gap in understanding these emerging dynamics as they relate to digital disruption in Australian farming, we employ a multi-level perspective on transitions in the Australian grains industry, combined with social practice theory to further explore social responses to smart farming technologies.The aim of our research is to explore the socio-technical factors that influenced stakeholder expectations about Big Data applications.Our key overarching research questions were: (i) what are the socio-technical conditions that assist the effective and acceptable use of digital agriculture and Big Data applications; and (ii) how might these technologies enhance or disrupt existing social and economic relationships in the agriculture sector?We examined the socio-technical factors that influence the impacts of Big Data applications at multiple levels, from the micro (the farmer and the technology developer) to the meso and macro (regulatory settings and institutional arrangements).

Two perspectives on understanding change: multi-level perspective on socio-technical transitions (MLP) and social practice theory
This research draws on the multi-level perspective on socio-technical transitions (MLP) (Geels, 2002(Geels, , 2012;;Schot and Geels, 2008;Geels, 2004), combined with social practice theory (Hinrichs, 2014;Shove and Walker, 2010;Ingram, 2015) to inform our analysis of stakeholder expectations of Big Data applications.The MLP approach was designed to provide "analytical and heuristic concepts to understand the complex dynamics of sociotechnical change" (Geels, 2002(Geels, : 1259)), making it a useful framework for understanding socio-technical factors related to emerging technologies.The MLP approach identifies three nested hierarchical levels of a socio-technical system: niche innovations (micro level); regimes (meso level) and landscapes (macro level), as illustrated in Fig. 1.
The socio-technical landscape represents the overarching level.It represents the wider context of macro, long-term economic, political, cultural and environmental trends and material context (e.g.physical infrastructure, population growth, economic development, resource availability, political ideologies and dynamics, societal values, climate change) (Geels, 2002(Geels, , 2004)).The socio-technical regime level refers to the dominant and relatively stable systems of interacting practices, social structures and institutional elements (e.g.cognitive routines, shared belief systems and expectations, as well as normative, regulative and formal rules) that shape the activities of relevant actor groups (Geels, 2002(Geels, , 2004)).Micro-level niches are protected spaces (e.g.R&D laboratories, demonstration sites) where radical novelties (innovations) are generated, incubated and developed.
The MLP approach includes a focus on socio-technical transitions, defined as major shifts in structures, cultures and practices such that the way societal functions are fulfilled is profoundly altered (Geels, 2002;Ingram, 2015).The MLP has been applied to a variety of contexts, including agro-ecological innovations (Wigboldus et al., 2016), algae blooms (Diaz et al., 2013), low-carbon transitions (Geels, 2012), transport systems (Nykvist and Whitmarsh, 2008), the UK Carbon Trust (Kern, 2012) and urban water systems (Quezada et al., 2016).As illustrated in Fig. 2, the central notion of the MLP approach to sociotechnical transitions is that these are non-linear, co-evolutionary processes that result from the interplay of multiple developments across these three analytical levels: "(a) niche innovations build up internal momentum, (b) changes at the landscape level create pressure on the regime, (c) destabilisation of the regime creates windows of opportunity for niche innovations" (Schot and Geels, 2008: 545).The MLP approach therefore helps make sense of the complex socio-technical dynamics and processes that must align in order for novel technologies to successfully disrupt the existing regime.For the purposes of this study we approximate digital technology related experimentation and uptake on individual farms or local, place-based networks of farms, as the empirical site of an emerging innovation niche.However, similar to Ingram (2015) in relation to the importance of niche-regime interactions, we recognise that there are multiple actors engaged with farmers through supply and information chains (e.g.scientists, advisors, retailers, cooperatives, industry organisations etc.) and that, in our conceptualisation, these actors directly and indirectly connect these innovation niches with regime-level structures and norms, situated within the broader socio-technical landscape.
Despite its utility, there has been constructive discussion in the literature on some of the limitations of MLP in examining the social dimensions of those transitions (Hinrichs, 2014;Shove and Walker, 2010;Ingram, 2015;Wigboldus et al., 2016;Raven et al., 2011). Hinrichs, (2014) in particular, in exploring transitions to sustainability in food systems, has argued that social practice theory can complement a multilevel perspective, and when used together can improve our analytical and explanatory purchase on food systems change.Social practice theory is a type of cultural theory that draws on the work of Bourdieu (1977Bourdieu ( , 1990)); Giddens (1984); Heidegger (1962) and Wittgenstein (see Schatzki, 1996) to explore how social structures and individual agency interact (for a comprehensive review see Reckwitz, (2002).While there is no unified approach within social practice theory (Schatzki, 2001), there is a common focus on understanding social life through the everyday and routine performance of social practices (Hargreaves, 2011;Spaargaren, 2011), including, for example, agriculture. Reckwitz (2002: 249) defines a practice as 'a routinized type of behaviour' made up of several interconnected elements, including both physical and mental activities, 'things' and their use, background knowledge and know-how and emotions and motivations.Thus, a practice is a 'rountinized way in which bodies are moved, objects are handled, subjects are treated, things are described and the world is understood' (Reckwitz, 2002: 250).
Shove and colleagues (Shove et al., 2012: 14) provide a useful account of three elements that make up practices, namely: (i) materials, including things, technologies, physical entities and objects; (ii) Fig. 1.Multiple levels (niches, regime, landscape) of a socio-technical system form a nested hierarchy (Nykvist and Whitmarsh 2008: 1375, adapted from Geels, 2002).competences, including skills, know-how and techniques; and (iii) meanings, including symbolic meanings, images, ideas and aspirations.The connections between these elements shape the way in which practices emerge, remain, change and disappear (Shove et al., 2012).Previous work by Hargreaves (2011) has found these connections to create subtle but influential shifts in the performance of working practices and the interactions and identities that sustain those practices.Similarly, Spaargaren (2011) argues that social practice theory provides new and useful ways of understanding issues of agency, technology and culture, which can inform environmental governance around sustainable consumption.Thus, social practice theory provides a valuable lens for understanding stability and change, including ways in which practices are established, reproduced or broken across space and time (Hargreaves 2011;Shove et al., 2012).Indeed recent scholarship has examined the ways in which smart farming technologies are accommodated within and modify the everyday lives, practices and identities of farmers (Carolan, 2016;Higgins et al., 2017).Following Hinrichs (2014: 143) we use social practice theory to recognise the importance and influence of "normal everyday routines and practices", the "possibilities of shifting these (or not)" and their relationship to politics, governance, values and ethics.That is, we focus on those social and institutional factors that influence uptake and outcomes associated with technology.These include actors' experiences with and perceptions of the risks and benefits of the technology and the extent to which the technology is seen to be compatible with existing farming or industry practices, routines and relationships (Rogers, 1995;Pannell et al., 2006;Vanclay and Lawrence, 1994).

Methodology
Our research adopted a broadly interpretivist, qualitative approach, following a social constructivist grounded theory strategy for analysis, informed by sensitising concepts described above from MLP and social practice theory (Charmaz, 2006).This allowed us to identify and explore different perceptions of, and experiences with Smart Farming and Big Data applications in the Australian grains industry.Through indepth interviews with diverse public, private and non-government organisations in the grain farming industry, as well as addressing our overarching research questions we specifically sought to elicit: (i) Multiple contexts of use of the technologies, e.g.how do Big Data applications integrate (or not) with existing practices and emerging trends in the agricultural sector?This question informs our first research question: what are the socio-technical conditions that assist the effective and acceptable use of digital agriculture and Big Data applications?
(ii) Characteristics of the multiple stakeholders and end users and Fig. 2. Multi-level perspective on socio-technical transitions (Schot and Geels, 2008: 546).
multiple technologies, e.g.how do different stakeholders understand Big Data applications and their associated risks and benefits?This question informs our second research question: how might these technologies enhance or disrupt existing social and economic relationships in the agriculture sector?Since the applications of Big Data are likely to vary across different agricultural sectors, we chose to focus on one of Australia's largest agricultural industries where Big Data applications are emerging, namely, the grains sector.The grains sector makes an important contribution to the Australian economy, with the production of grains, oilseeds and pulse crops accounting for around 23 per cent ($13 billion) of the total gross value of farm production and around 24 per cent of the total value of farm export income in 2015-16 (ABARES, 2017)).Grain production (predominantly wheat) occurs across a wide range of distinct agroecological zones, each with different climate and soil characteristics (Guthrie et al., 2017).Family farm ownership continues to dominate Australian grain production with more than 95% of grain farms being family owned and operated (Kalisch Gordon, 2016).The Australian grains sector is an example of an industry that has had to embrace niche innovation at the farm level in the past in order for farms to remain viable in an unprotected market and to mitigate the risk of significant climatic variability.Innovation examples include the uptake of no-till farming, GPS technologies for auto-steer and, more recently, modest rates of adoption of precision agriculture approaches, such as variable rate fertiliser application (Robertson et al., 2012).Practices can change rapidly in order for grain farmers to remain competitive and as such it provides an interesting case study industry with a history of niche innovations taking off.Examples of smart farming platforms in the Australian grains industry include ProductionWise (GrainGrowers, 2018), YieldProphet (BCG, 2017), Graincast (CSIRO, 2018) and Smart Farmer (Elders, 2018). 1  With the help of subject matter experts known to the researchers, we identified the key stakeholders in the grains industry.We selected interviewees using a purposive sampling approach (Patton, 1990), collaborating with key informants to identify and recruit participants from different sectors within the grains industry and with different levels of involvement with digital technologies.Participants were invited via email and a follow-up phone-call.We conducted semi-structured interviews with 26 grains industry participants (23 men and 3 women): 14 participants were from the dryland broad acre and mixed farming systems of the Wimmera-Mallee region (in the southern state of Victoria), providing regional level grains sector insights; and 12 participants represented a cross-section of other industry stakeholders from the state and national level (see Table 1; see also (Jakku et al., 2016); Ethics clearance #070/15).The Wimmera-Mallee region was selected because the region is engaged in active discussions about the future of digital agriculture.At the time of the interviews a key local grower group with a strong innovation record -the Birchip Cropping Group, a farmer-led agricultural research and extension organisation -was exploring opportunities for setting up a data co-operative, while the local council was developing a digital futures strategy.In the Wimmera-Mallee Region there is approximately 3 million hectares of dryland cropping and livestock country predominately cropped (approximately 75%).There are approximately 4000 family farms in the region and they make up the vast majority of landholders.Farm size ranges from very small (under 200 ha) to over 5000 ha, with some up to 10000 ha.The Birchip Cropping Group has 430 family farms as members, farming approximately 1 million hectares.A typical membership consists of two family units farming the land together.Often both families will have children back on the farm or expecting to be back on the farm shortly.Farm size is mostly commonly between 2000 to 4000 ha.
Interviews were generally one hour in duration and nine were conducted face-to-face in the Wimmera-Mallee region in February 2016, while the remaining 17 interviews were conducted via telephone between January -March 2016.The interviews started by covering some background information on individual participants.This was followed by questions about their place within the grains industry supply chain and their views on information flows and relationships among key players.Next, the interviews explored perceptions of digital agriculture and Big Data, prompted by the following questions: (i) what does Big Data and digital agriculture mean to you?; (ii) how much is Big Data part of your current business or future strategy?; (iii) what benefits or opportunities do these digital technologies and Big Data applications provide?; and (iv) what problems or risks do they present?The final section explored ideas about how these risks might be managed or reduced and some final reflections on the future of digital technologies and Big Data in the grains industry.
The interviews were audio recorded and transcribed.We used the qualitative data analysis software QSR NVivo® (QSR International, version 10) to aid the coding, analysis and management of the data.Interview transcripts were analysed using a mixture of deductive and inductive coding followed by iterative thematic analysis (Grbich, 2007).This means that our research questions and understanding of social practice theory and MLP guided us in determining what codes might be relevant, which were then adapted and shaped from what was actually found in the data.The result is a hierarchical structure of themes, sub-themes and nodes through multiple rounds of coding, informed by (and informing) our theoretical perspectives (see appendix for the final coding structure).
In the results section that follows we explore key themes that emerged in the interviews regarding stakeholder perspectives on the benefits and risks of these smart farming technologies.Then in the discussion we return to the two theoretical perspectives -the MLP approach and social practice theory -to extract key interpretive insights that these complementary perspectives offer.

Stakeholder perspectives on the benefits of smart farming and Big Data
The grains industry stakeholders that we interviewed identified a range of benefits and risks associated with smart farming and Big Data, as summarised in Table 2.

On-farm benefits
On-farm benefits were the most frequently mentioned type of benefits associated with emerging Big Data applications.Interviewees outlined a range of ways in which Big Data could improve farm management and decision-making, focusing on improved efficiencies through more targeted applications of on-farm inputs: Well, benefits are more targeted application of inputs across our farms, so if we have the data to be able to aid in decision making then we can match our inputs to the potentials of the season.Not only on a paddock scale but down to a pixel scale or a particular point on the farm.(Grower group 8) A related theme was the increased productivity and profitability that improved farm management and decision-making could bring: "… farmers make lots of decisions through the life of their crop on a farm, and if they can make better decisions…they are able to improve their efficiency and productivity out of that" (Input provider 3).The value of real-time information for decision-making was also highlighted as a potential benefit associated with advances in digital agriculture, especially sensing technologies: "So having a system -it might only be one or two weather stations on a property -that feeds in almost real time data to the farmer; would be really useful" (Grower group 6).Future benefits of Big Data were expected from linking-up current or future data sets, such as soils, climate projections, weather forecasts, water models and crop information on an individual farm basis.
Several participants expressed views that the efficiencies enabled by Big Data would mean higher prices for growers at the farm gate, although this view was not always supported by growers themselves: So digital agriculture, for me, is the automation of a lot of management processes for farmers largely, I don't think that you could necessarily automate much more of the supply chain.So it…is the next productivity gain pre-farm gate, it means that growers more take a role of managing their farm and putting prescriptions in place to automate those processes, [such as] a sprayer going out and spraying without a human actually sitting in it… (Grower group 4)

Industry and supply chain benefits
Interviewees also identified a number of potential industry and supply chain benefits, particularly those related to optimisation along the supply chain and the improved industry-level decision-making that this could bring: "I think, the supply chain will streamline itself and will be able to drive efficiencies from the use of Big Data, so there'll be a commercial benefit for the business" (Grower group 4).Another important benefit associated with Big Data was improved predictive and analytical capabilities for storage and transport logistics providers: "Certainly through logistics…being able to track, maintain, record is important for supply chain optimisation.…So at every level it will drive improved performance" (Grower group 3).
More accurate tracking and predictions of yield would allow for better optimisation of decisions and resource allocation related to transport, logistics, labour, timing and price points (Sonka, 2016).One interviewee referred to this as visibility along the supply chain where there may have been unknowns before, such as when grain was transported or combined with other growers' grain: The marketers want to keep data separate to get commercial advantage whereas [from a logistics and handlers' supply chain perspective]…greater visibility and accuracy around that data is what we're chasing.(Logistics and trading 2) Big Data applications could also increase the traceability of grain in the supply chain, creating value for consumers, retailers and processors as well as growers: Traceability is the one that everyone talks about, so traceability is a good example because we're seeing increasing demand for people who understand where their food came from.The digital technology will enable that.(Input provider 1) Similarly, one grower described how information on varieties of grain (including GM varieties, provenance, quality) could now be traced by customers, creating premium products and niche markets with potential to grow demand for specialist products (e.g. grain for craft brewing).
Improved crop forecasting data was reported as another potential future benefit of Big Data.Interviewees noted that the ability to gather and analyse data on variety, quantity, location, quality, weather events, management decisions and market prices offers a whole new way of understanding the grains industry.However, support will be required to improve the capacity to interpret the data in order to answer specific questions, for example to compare years and management decisions, as well as to look at non-traditional indicators to open new market niches.
Bulk handlers described the potential for value generation from collaboration between companies involved in grain storage and transport.One handler described the benefits of data-driven predictive modelling of the location, timing, volumes and quality of grain yields for informing decisions on rolling stock and road transport.The efficiency of the system overall, including managing grain flows into the port terminals, could be improved by co-investment across grain handling companies in a given region in generating a 'complete' picture of where and when grain was moving.However, investment in such a system by one company alone was seen as unviable and therefore improved collaboration across supply chains would be necessary to fully realise the potential benefits of Big Data applications.

Table 2
Summary of results on perceived benefits and risks of smart farming and Big Data applications.

Benefits Risks
On-farm • Improved efficiency (e.g. through more targeted application of on-farm inputs and automation) • Increased productivity and profitability • Real-time information to help make better decisions • Linking data sets to create greater insight Technical • Novel and immature technology • Concerns about data accuracy, reliability and transferability • Challenges of data storage and handling • Challenge of interoperability • Agricultural data fragmented • New skills and capabilities needed • Limitations of digital infrastructure Industry and supply chain • Optimisation along the supply chain • Improved industry decision-making due to more accurate tracking and prediction of (collective) yield • Improved predictive and analytical capabilities for storage, transport and marketing logistics • Traceability and opportunities for premium products and niche markets

Social and institutional
• Concerns about data privacy and security • Uncertainty over principles, rights and compliance regarding data sharing, ownership and use • Power asymmetry within industry • Concerns and lack of trust regarding third party (corporate) use of and profit from on-farm data • International competition • Value proposition for sharing on-farm data not clear for many farmers

Stakeholder perspectives on the risks associated with smart farming and Big Data
Despite the potential benefits of Big Data applications, interviewees identified a range of concerns about these emerging technologies, from technical concerns about the technologies to broader concerns about the social and institutional context within which these new technologies are operating.

Data accuracy, reliability and capabilities
One common theme that emerged was that Big Data applications are novel and immature technologies.The 'teething problems 'often associated with new technologies, combined with concerns about data accuracy, made people wary about the reliability of emerging Big Data applications: But given this is…a relatively new field, it is going to take some time to get that validation and to get the systems working at a high level of accuracy.So that's, I think, one of the challenges over time that farmers are going to need to be able to work with systems that might not be perfect, but as they work with them, they will get better.(Input provider 3) A related concern here was the challenges of data storage and handling in the context of a new and emerging technology: So we've got a lot of data that we've been collecting.And where we can use it, we do.But we really haven't found an easy outlet for that.And I think that's one of the things that the Big Data problem's created.There's just lots of information, which we know we need to collect to be able to get enough to be useful, but we don't know what to do with it and we should have enough by now to be useful.(Grower 3) The transferability and applicability of the new technologies were another related area of concern, including the difficulty of making judgments about competing technologies in this domain: …there's at least half a dozen companies [in the United States] offering precision, prescription farming services for farmers to deal with, typically nitrogen in corn.…So those sorts of services are going to come here eventually, and how does a farmer evaluate whether the Pioneer solution is better than the Monsanto solu-tion…? (Grower 1) Interviewees also noted that data within the current Australian agricultural system is highly fragmented and people are not currently maximising existing data.A related issue here is the challenge of interoperability (the ability of information technology systems to exchange and make use of information), which is compounded by the fragmented nature of Australia's digital agriculture data landscape: …it's like different railway systems.In the end, it's sometimes easier to do it your own way than find a compromise.And I think that goes back to trust and everything like that.It's how much are you willing to give up and how much are you willing to drive forward?(Research & consulting 3) Furthermore, it takes new skills and capabilities to properly apply and interpret the results of Big Data systems and analysis, as identified by this grower: It will change the skills, over time, required to be a successful farmer.(Grower 1)

Digital infrastructure
The fundamental limitations of Australia's digital infrastructure, especially in rural and regional areas, was one of the most significant barriers to the success of Big Data applications identified by interviewees.There were widespread views amongst interviewees that the mobile phone network and internet access in rural Australia was not currently sufficient to support some of the potential advances in Smart Farming: …another risk is actually not having the ability to download all this data and actually upload stuff and have good internet coverage.…If we want this to happen in the country we've got to have our mobile phone working pretty much, and that's a major concern.(Grower 2) Although one interviewee acknowledged that "there are work arounds" for the limitations to digital infrastructure (Grower 1), there remained a degree of scepticism about how much the rural and regional digital infrastructure would improve in the near future, which some interviewees linked to the further widening of the city/country divide: And our other big problem that's going to become more pronounced is just lack of Internet access… I'm not sure what's going to come out of the rollout of the NBN [National Broadband Network] but…we're going to be left off the end of that and a lot of this sort of stuff is going to be quite data hungry that we should and could be using.So I'm not sure what the answer to that is but it's certainly going to create a bigger city/country divide.(Grower 5)

Governance of data privacy, security and ownership
The adequacy of current regulations and practices to protect the privacy and security of farmers' data was another issue that was mentioned by several participants.Some interviewees were satisfied that privacy and security measures would be adequate: So we have privacy policies that are inserted into our licensing agreements with growers on an annual basis.And they obviously take into account federal and state requirements.And we update them as there are any changes in local requirements in Australia around privacy.(Input provider 3) However, other interviewees expressed more concern: …All that privacy stuff, it's just can of worms.And it's got the potential to completely explode.But we are so reliant on our technology nowadays, that we can't really stop it.…So definitely some healthy scepticism and concern about how that sort of privacy can go.And I think people just need to become more and more aware of it -me included.And making those safeguards to make sure that you can protect your data.(Grower group 6) Moreover, even with privacy and security measures in place, breaches are always going to be a risk: …I think that it is incumbent upon organisations that are storing data that they need to be doing their very best to maintain that security, but at the same time the consumers and the farmers need to understand that there can be breaches that happen from time to time even with the best intentions.And that's always going to be a risk.(Input provider 3) The broader dynamics surrounding data ownership, data sharing and the way in which the benefits from this are distributed also emerged as key themes.Data was understood as a valuable commodity, hence data ownership was an important theme within discussion about the implications of these emerging technologies: "I have to admit everyone seems caught up in [the idea] that data will be valuable therefore I should focus on owning it and extracting insight from it" (Input provider 1).Many growers were concerned that large corporations could capture this value, possibly at the expense of local growers, based in part on observations about how these issues are playing out in the United States: Now, if there is value in it you kind of want to make sure that if we're doing all this, we want a little bit of something back and I guess the fear is the big players swoop down, grab it, run off and make some big business model and they make a good living off it and the guys that generate it all miss out.…I'm of the thinking that we're probably at the bottom of the food chain.We've got something that maybe someone wants collectively and if they get it for nothing it just doesn't feel right.(Grower 4) Many interviewees expressed uncertainty about rights to data ownership and use: It would be…really good to know how the information could be misused…actually that's probably as relevant as anything, to be honest, the risk side of it, how could it be misused, so then we can make an informed decision about where it goes and how it's used.(Grower 2) Competing views were particularly evident regarding the principle that growers own their on-farm data.Some interviewees accepted digital agriculture service providers' assurances that growers' retained ownership of their on-farm data under emerging Big Data opportunities.For example, one interviewee discussed the approach taken by a major company to ownership of farm data: But effectively what [the company] has said is that if data is generated by a farmer or from their farming equipment on their farm, that data is owned by the farmer.… And…if a farmer brings their data to [the company] or they generate it through their equipment and it's used as part of [the company's] systems, if the farmer wants to leave, they can take their data with them, and we [the company] don't own that data.…That's why I do really like the guiding principles that [the company] have put in place.(Input provider 3) However, other interviewees expressed a more critical view, focusing instead on how that data may be used and by whom it might be used.Thus, lack of trust in data ownership regulations is a key factor mediating perceptions of these new forms of technologies: They [digital agriculture service providers] say the farmer owns the data, the farmer, legally that's true but practically what does it mean?Almost nothing.A far more interesting and pertinent question is what are they doing with that farm data?(Grower 2) The lack of trust also reflects established belief systems and normative roles between farmers and agribusiness.Several participants referenced the unequal power relations that were seen to exist between farmers and large businesses: But it depends on how the information is going to be disseminated once it's collected as well and who has control of it.And that's one of the areas that really worries me is that it seems to me that most farmers are still reasonably small and most of the businesses they deal with are reasonably large so there's going to be an inequity in the data.(Local government 1) Concerns around data ownership and use also related to the boundaries around what data growers are comfortable with sharing and what data they want to protect: I'd be more worried about that…when you sign the dotted line to buy that tractor you lose control of the data without really realising it.… Well I think one farmer versus [a large company], we've got Buckley's.…[T]he information about how many hours our tractors do, what sort of conditions do they work in, what problems they have, that's all great information to have.[The large company] needs that information to build better tractors, more efficient, which is going to benefit us.I think the data that we have that's most value to other people is our yields, our varieties that are much more specific to our farms.Our gross margins, our business information.And that we should be more able to keep control of.(Grower group 6) As a consequence, some interviewees raised doubts about the willingness of growers to share their on-farm data, even with other growers: But I can't see people openly sharing their data.I can see people giving you a bit of something that you might need, or sitting down with your agronomist, giving them some stuff.But I can't see me and you, being farmers that are 100 km away, really, openly sharing.We might talk on basics, but I'm never going to let you take my yield maps and you're never let me take yours.We might look at them together and talk about different farming methods and the physical, but you're never going to walk away with that data, I wouldn't have thought.(Input provider 4) It was clear that the perceived adequacy of data ownership and third party benefit regulation, along with inequalities and lack of trust between farmers and large agribusinesses were directly connected with willingness to adopt in practice Big Data applications that involve data sharing: "I think the risk lies in farmers being confident that they don't need to lock up their data and make it absolutely unavailable to anyone except a very narrow limited range of providers" (Grower group 2).One interviewee pointed out that the industry needed to better explain the value proposition for access and use of on-farm data: …the industry has done, frankly, a terrible job of explaining why they want access to farm data.Not so much an issue probably here in Australia yet, it's probably just starting to happen now, but in the US it's been going on for quite a few years, and it's even more so.….So it's this weird thing where they don't want to tell us exactly what they're doing with it but if they don't tell us what they do with it, why would we trust them?…[P]robably mostly they are doing the right thing but that's not explained anywhere and we're certainly just trusting that's what they say they're doing, there's no way of verifying it…and that's what's holding more farmers back from adopting it, but we miss out on the benefits of it then as well.(Grower 1) Thus, issues of trust and transparency, based on normative roles between farmers and agribusinesses, have the potential to constrain the willingness of farmers to participate in smart farming technologies.

Discussion
Our interviews reveal how grains industry stakeholders portrayed the potential benefits that smart farming technologies offer, both onfarm and across the industry value chain.Interviewees also raised a range of concerns about these emerging technologies, from technical issues such as data accuracy and reliability through to broader concerns about the social and institutional context within which these new technologies operate and are governed.We now consider how theoretical insights from the MLP approach and social practice theory can inform our interpretation of grains industry stakeholders perceptions of these benefits and risks associated with Smart Farming and Big Data applications.
The MLP approach identifies three nested hierarchical levels of a socio-technical system that socio-technical transitions must navigate: niche innovations (micro level); regimes (meso level) and landscapes (macro level).Drawing on the MLP approach, we found that niche level mobilisation processes are currently using on-farm benefits as the key means to establishing a vision and value-proposition for smart farming.Smart farming platforms in the grains industry such as ProductionWise and YieldProphet are designed to help improve on-farm decisionmaking.The fact that these benefits were also expected to accrue across the supply chain and even more broadly is likely to facilitate changes to processes and systems at the regime level.This confluence of benefits could be harnessed to support the collaboration required for effective use of Big Data applications, if potential barriers and risks of the kind identified in our study can be mitigated or managed.For example, the potential to derive national benefit from Big Data applications may support collaboration across firms and organizations along with regional co-investment, which participants believed would be needed to support the diffusion of such innovation.This supports commentary in the literature that collaboration across firms and organizations will be necessary in order to fully realise the potential benefits of Big Data applications in agriculture (Sonka, 2016).
However, elements of the socio-technical regime and landscape are constraining niche level innovation.For instance, a major constraint at the landscape level is the current state of Australia's digital infrastructure, which in rural and regional areas is not currently sufficient to support the full potential of smart farming technologies.The challenges surrounding rural and regional digital infrastructure and the growing divide in data infrastructure quality between urban and rural areas have been noted elsewhere and are a significant issue to overcome in the effort to build sustainable digital futures for rural and regional communities (Salemink et al., 2017;Roberts et al., 2017;Pant and Hambly Odame, 2017).
There are also important niche-regime interactions shaping social responses to smart farming.The current state of data governance at the regime level is an important factor, as seen by the range of concerns held by stakeholders about the adequacy of current regulations to protect the privacy and security of farmers' data and to manage data ownership rights (see also Dyer, 2016).There are currently no governing legal principles in Australia that clarify and build trust for producers around the access and use of agricultural data.This lack of agdata standards and licensing arrangements contributes to the lack of trust that producers have towards data contracts (Wiseman and Sanderson, 2017).Digital agriculture service providers seek to address concerns about privacy and data ownership through written contracts, which specify the terms and conditions regarding data ownership and use (Keogh and Henry, 2016).The lack of trust is in part due to the way in which some user agreements "bury exclusions deep in the document which in effect give free reign to the software providers…to use the data in many different ways, including via the sale or transfer of the data to a third party" (Keogh and Henry, 2016: 37).Lack of trust due in part to previous experience, can lead to apathy and withdrawal (Stern and Baird, 2015).Other researchers have identified that concerns over data sharing are also related to the dynamics of power relations between industry stakeholders (Wolfert et al., 2017;Avelino and Wittmayer, 2015;Nelson and Tallontire, 2014;Avelino and Rotmans, 2009).Thus, institutional arrangements for data ownership and sharing in the current regime are not providing a trusted environment necessary to encourage a willingness to share data.
Although the MLP approach provides a useful lens for examining the multi-level dynamics shaping potential trajectories for Big Data applications, we now turn to social practice theory for further insights into how people are currently responding to these emerging technologies.While the MLP approach puts technology at the centre of analysis, social practice theory focuses on the actions of people and the way in which images and meanings, materials (including technologies) and competencies and skills shape dynamic systems through interacting and co-evolving practices (Shove and Walker, 2010;Shove et al., 2012).Social practice theory therefore helps to uncover the importance of human agency and sheds light on the "inevitable contests and politics" that are involved in transitions (Hinrichs, 2014: 149).
In terms of the images and meaning given to smart farming technologies, grains industry stakeholders that we interviewed considered Big Data applications to be one of the most important developments in agriculture.These new technologies were portrayed as offering the potential to transform Australian agriculture through improved on-farm decision-making, prediction and analysis, leading to significant productivity gains.On the one hand, this is part of an emerging shift in what it means to be a farmer, with increasing emphasis on management skills and knowledge over hands-on labour.On the other hand, these smart farming technologies build on existing precision agriculture technologies and are therefore broadly compatible with existing everyday routines and practices within dominant forms of farming.However, we found that perceptions of who would benefit most from these emerging technologies shaped the images and meaning given to these technologies.Furthermore, data sharing arrangements are not currently part of everyday routines and practices, which underpins some of the concerns around putting Big Data applications into practice.
While the potential for Big Data applications to aid decision-making at both the on-farm and broader industry level was a common theme in descriptions of the benefits of these technologies, the greatest financial returns on implementing Big Data approaches were largely reported to be tied to businesses upstream and downstream of the farm gate (i.e.input suppliers and manufacturers, traders and marketers), rather than farmers themselves (see also Fleming et al., 2018).Marketers and traders expected that Big Data would allow them to better predict export demand and prices, however growers expressed concerns that this could exacerbate the commercial advantage these groups currently exercised over growers.Consequently, growers were concerned that disproportionate benefit from Big Data applications would accrue to businesses upstream and downstream of the farm gate and indeed believed that growers were most exposed to potential risks and exploitation of these technologies within supply chains already characterised by power asymmetries.Furthermore, many participants also made reference to the city/country divide.This deep cultural pattern was linked to issues of trust and inequality more generally, such as the belief that the benefits of Big Data would accrue to large corporations (urban entities) with farmers (rural actors) losing control of their own data and thus the benefits to be derived from it.This aspect of Australia's cultural identity is influencing how Big Data applications and related innovations are perceived.This is further exacerbated by the lack of trust in current data governance arrangements, outlined in relation to the MLP approach.When combined, these factors shape perceptions of the meaning and value of these technologies and their potential implications for stakeholder relationships in the agricultural sector.
Social practice theory also focuses attention on skills and competencies that are needed to underpin novel practices.It takes new skills to properly apply and interpret the results of Big Data systems and analysis.However, the skills and capacity of different grains stakeholders to engage with or benefit from Smart Farming (including Big Data applications) is presently highly varied, especially amongst farm businesses.There were growers who were investing heavily in digital technology, sensing, automation and other data intensive elements of their business in readiness for future Big Data applications.These growers reported that they were already realising benefit from that investment.However, for the majority of growers we interviewed, the benefits they receive are likely to be realised longer term or 'down the track' and in many cases appear uncertain.In the short to medium term, concerns about transparency, equity (in terms of the distribution of benefits), data ownership and access appear to dominate.Similar themes have emerged in other studies investigating digital transformation in the farming sector in Australia (Guthrie et al., 2017;Zhang et al., 2017) as well as in New Zealand (Shepherd et al., 2018), North America (Bronson and Knezevic, 2016;Carolan, 2016;Carbonell, 2016) and Europe (Wolfert et al., 2017;Poppe et al., 2015;Regan et al., 2018).This suggests that our findings have broader applicability, particularly in terms of the importance of trust and transparency as central themes underlying social responses to the risks and costs associated with the use of on-farm data, which in turn have the potential to limit the informed and consensual participation of all stakeholders in Smart Farming and Big Data applications.

Conclusion and recommendations
The successful implementation of Smart Farming and Big Data applications depends on ensuring that the design and implementation of these technologies respond to stakeholder dynamics within the agricultural sector, including the way in which these novel technologies are understood, adopted and adapted in practice by farmers and other decision-makers (Sonka, 2016;Wolfert et al., 2017;Bronson and Knezevic, 2016).Trusted information and advice networks are likely to be important mechanisms for growers in mediating the benefits and risks of engaging with these regime factors.As such, alignment of these new opportunities with existing (or re-negotiated) trust relationships is a critical enabling condition.Therefore, our primary recommendation is the need to invest in building the capability of growers and farm businesses to be both informed data consumers as well as co-creators and curators of data, by involving growers and their trusted information and advisory networks in the cooperative development and trialling of these systems.We argue such actions would broker the understanding of everyday practices and decisions of farmers at the niche level, with the networks, norms and structures of regime-level elements that enable or constrain possible transitions.
Key questions remain, for example: what are the implications of emerging and diverse models of services and governance of Smart Farming and Big Data applications, and how adequate are they at meeting the requirements of transparency, shared benefit and access raised in this study?We found that participants' views were divided on this point.Furthermore, since advances in Smart Farming are also likely to converge with a variety of information-based compliance processes (e.g.food safety and environmental regulation), it will also be necessary to understand how this regime level convergence is likely to impact on farm productivity related developments and the regulation of agricultural production and supply chains at the enterprise level.
There is an important role for social researchers working in a participatory way with industry, corporate, research, development and extension and other stakeholders to identify the complex and intertwined factors and processes influencing the deployment of Smart Farming and Big Data applications.Such research will support their ongoing improvement, assess their transferability between sectors or growing regions and ultimately help to ensure that the application of these novel technologies has the widest possible benefit across the agricultural sector.

Table 1
Interview participants by stakeholder category.
1 ProductionWise and YieldProphet were both available at the time of interviews and mentioned by various interviewees.Graincast and Smart Farmer have both been recently released (after the interviews were conducted).