While these may be useful in providing novice farmers with a starting point as to how to even begin putting together information regarding environmental, resource, and operational data, most of the farmers we spoke with end up restructuring their previously collected data to fit policy and/or certification requirements.The population of farm workers includes a large Spanish speaking population and a diverse range of literacy levels, as well as a varied set of communication needs across stakeholders. Such diversity is not always accommodated. Models may include multiple languages or a visual focus to overcome barriers to communication in the social context of the farm.There is a lack of consistency and technological capacity across the spectrum of sustainable agriculture as a result of the sparse attention that has been paid to the information management needs and requirements of sustainable farmers. Their agricultural systems cannot be adequately represented and analyzed using current modeling tools and methods, and many environmental data sources have not been designed to integrate with the tool kits that such farmers currently use.Artifact trails spanned across disparate data structures: from hand-drawn maps containing crop locations, rolling benches spreadsheets varying in structure from year to year even within a farm, to custom databases to handle portions of the farmer’s workflow.
Maintaining coherence and consistency across data structures is key to stitching together the information workflow of sustainable farmers; the heterogeneity of data structures within farms provides a significant stumbling block in effective workflows. One potential avenue for future work involves the design of better models to encapsulate the types of data that are collected and fit with the data collection techniques that farmers engage in. Models are apt devices for communication of a system’s composition and performance, as well as useful artifacts for reflection. In particular, a domain-specific modeling language would enable system stakeholders to gain perspective on their systems, perform environmental analyses, and create abstract yet grounded models that they can manipulate before changing the real world systems that are represented.The switching between data structures is accompanied by several transitions between different mediums of technology, resulting in a messy information management experience. Five genres of technology were used across one or more farms: physical or paper-based artifacts; communication technologies like email; external regulation- or certification-specific software; farm-specific custom software; and office productivity software such as Microsoft Word and Excel.Farm models varied in level of abstraction, formality, and granularity, both within an individual farm and across agricultural system types. For example, on some farms, resource use was tracked at the whole farm level, while on others there were intricate interconnections between subsystems requiring tracking at a granular level.
The commodity and process complexity of the farm also affected the level of abstraction, formality, and granularity of models. Models need to be flexible enough to capture whole-farm activities as well as fine-grained data about specific farm components. Ideally, farmers should be able to create both informal and formal models depending on the type of data they have and analyses they plan to conduct. Consequentially, relevant tooling must be capable of representing the varying spatial and temporal complexity present throughout the spectrum of sustainable agriculture.Many of the models used for internal assessments by farmers are created in an ad hoc manner and for a specific purpose For example, a map created for communication to visitors may be opportunistically used for coordination among farm workers. However, this is a one-off reuse, and is not inherently supported by the models. Representations cannot always be easily repurposed, resulting in a significant reusability gap.Further, data is often isolated in purpose-specific models. For example, input logs used to track resource application for organic certification are not necessarily connected to inventories that are used to track expenditure on materials. Data and effort are duplicated as data is tracked separately for inventory management and organic certification. Farm data needs to be captured in a general enough form such that it can be manipulated and transformed on demand. There is an opportunity to reduce the reusability gap by enabling farmers to create modular component-based farm models. There is also a significant amount of publicly available environmental data, such as data on soil composition throughout the United States, global weather data, and California-wide water quality and availability. These data can and should be incorporated into information management tools for sustainable farmers to reduce their data collection burden.The transience of agriculture means that not all data collected is necessarily archived. This is further exacerbated by the rate of change on farms: the more often the farm changes, the easier it needs to be to update the model. This also affects the formality and reusability of farm models. Use of model configuration management may allow for farmers to track changes in their farm models just as they use models to track changes in their farms. This would also allow farmers to compare models over time and reduce the effort involved in updating models to reflect changes. We found workarounds implemented by farmers to mimic such a workflow. By designing with the intent to archive historical data and track changes, the inherent dynamism of the farm can be captured.There is a mismatch between the causes and goals of data collection, particularly as the farm evolves. While initial causes for data collection may be for regulatory reporting and system understanding, eventually farmer goals can expand to include communicating to diverse stakeholders , environmental analyses, and monitoring.
The mismatch between the collection trigger and emergent goals results in a lack of coherence in farm models and data workflows. We must explicitly consider both the causes and goals of data collection in the design of information management tools for sustainable farmers to ensure that appropriate data is aggregated and connected. Supporting easy transformation of the information management structures and practices would facilitate synchronicity among causes and goals.Farm stakeholders are constantly using, sharing, and communicating various data for daily decision-making and problem solving. However, many of the technologies we found in use were not designed to meet the coordination and collaboration needs of varying farm stakeholders. Any redesign of the information workflow of sustainable farmers must provide these stakeholders with the capacity to conduct environmental assessments and other forms of analysis; coordinate among farm workers and other stakeholders; and communicate with regulators, certifiers, and consumers as needed. Explicit attention to the differing privacy and access characteristics of data would be critical.They have been used in software design as stand-ins for real users, allowing people to engage in human-centered design, where the user experience and interaction of the human with the system is the key focus. In this dissertation, I flip the traditional persona, and instead use it to describe a system , that a human , is interacting with, cannabis drying room the result of which is a Farm Persona. These Farm Personas were created based on the data and findings of Chapters 3 and 4. Methodological detail on the design of these personas is available in Section 6.1 of Chapter 6, as the Farm Personas are explicitly presented during the evaluation of MoSS. The Farm Personas were used to engage in Persona Driven Design. Partial MoSS models were created to represent various activities, components, and aspects of hypothetical farms. These models were then used to refine the Pseudo-Software Models and thus refine MoSS. For example, Figure 5.8 below shows how a Farm Persona called Blackbird Gardens was used to explore the representation of crop layout and field management. Persona Driven Design thereby allowed for the farm-centered design of MoSS.Personas are detailed constructions of fictitious yet archetypal users of a software system that allow for human-centered design. There is growing interest in using personas to engage in design that considers other forms of system interaction. For example, collaboration personas have been used for the design and evaluation of tools for use by groups of humans. Non-human animal personas were proposed in recognition that the human stakeholder should not entirely dictate prototype development: a cow is a user of a robotic milking device, and the cow’s needs, welfare, and experience in mind, should dictate the design of the tool. Farmers are integral actors in agricultural systems. Often, in addition to being the farm’s owner/operator, a farmer is also: a manager of staff, the primary decision maker regarding on-farm activities, a stakeholder of farm data, and a farm data collector. While future work involves the design of a modeling tool to allow the farmer to interact with MoSS, the scope of this dissertation is to design a mechanism for modeling sustainable agricultural systems. I therefore consider farmers as primary actors in a System Persona. I define a System Persona as a detailed construction of a fictitious yet archetypal system. As the focus of this dissertation is on the holistic representation of the characteristics of small- to medium-scale sustainable agricultural systems, I created a set of Farm Personas, where each persona represents a hypothetical small- to medium-scale sustainable farm in California.The goal of this activity was to flesh out the Farm Personas with enough details so as to be used as part of the Persona Driven Design of MoSS, as described in Section 5.1 of Chapter 5. The Farm Personas needed to be grounded in both the academic literature regarding formal modeling in sustainable agriculture , as well as the findings regarding sustainable farms in reality . Overview tables were created to map out the key characteristics of the 16 farms from Chapter 4, Farms Atwood to Pullman . These led to the creation of design parameters for the MoSS Farm Personas. These parameters reflect greater nuance regarding the characteristics of the farms. Once the base Farm Personas were created, they were augmented through the addition of characteristics from the 16 farms resulting in artificial yet data-rich composite farms. Two primary characteristics were used to augment the Farm Personas to engage in the design work that resulted in the MoSS framework: complexity and dynamism. This process is briefly described next.A scenario is a hypothetical set of activities to be enacted by actors interacting with a system of interest in the real world. Scenarios have been used in software engineering and HCI research both as a means to explore potential designs of a system, as well system evaluation. For example, Sutcliffe describes how scenarios may be used in requirements engineering, in particular, to check abstract models through use of scenarios as a substitute for formal verification. Scenarios are often implemented when the problem domain is “squishy”, i.e., the problem boundaries are not distinct, the interactions are complex, and the problem does not lend itself well to linear design work or structured evaluations. It follows that there is evidence of the use of scenarios for design and evaluation in the environmental assessment and sustainable agriculture communities. In LCA research, scenarios have even been used to represent hypothetical agricultural systems to explore, for example, the environmental impacts of various pig production systems. The University of California Cooperative Extension has been conducting “Sample Cost of Production Studies” for various farm commodities since 1928. While the early reports are written in the style of a handbook, later reports in the mid-2000s begin to list study assumptions in the style of a scenario. For example, the 2009 cost production report for organic leaf lettuce in the central coast region of California, begins with a description of hypothetical farm with specific produc-tion activities; labor, interest and equipment; and how cash and capital are spent and obtained. It is these characteristics of a fictitious farm that provide context for the various cost estimates that are subsequently listed. This structure is powerful as it allows the reader insight to the rationale for listed costs, contextualizes the calculations, and provides a human readable example that a reader may work through. These cost production studies also proved an invaluable model for tailoring the software-style scenarios for sustainable agriculture.There are eight scenarios described in this chapter that were used to evaluate MoSS. The primary basis for scenario creation were the findings of the qualitative study involving 16 sustainability-oriented farms in California, described in depth in Chapter 4. To ensure scenario accuracy, interview data were also checked against relevant literature prior to being incorporated into a scenario. To avoid overspecialization and simply designing for those 16 farms , I created a set of scenarios, applicable to farms throughout a spectrum of sustainable agriculture.