Perspectives on Explanation Formats From Two Stakeholder Groups in Germany: Software Providers and Dairy Farmers

Mengisti Berihu Girmay, Felix Möhrle · 2025

Read Perspectives on Explanation Formats From Two Stakeholder Groups in Germany: Software Providers and Dairy Farmers by Mengisti Berihu Girmay, Felix Möhrle (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-11665.txt.

Introduction

Software that flags sick cows must explain itself to the farmers who act on its warnings. Mengisti Berihu Girmay and Felix Möhrle designed four explanation styles in 2025 for a mastitis alert, from plain text to herd comparisons and time charts. Farmers had rated the styles earlier; now thirteen software providers guessed how farmers would rate them.

The guesses missed in telling ways, backing the worry that builders assume what users want. The study is small and the authors say so plainly. Still, the gap points to money and trust lost when digital tools land badly on farms. They argue for asking users first, properly, before building the explanation in.

Catalog record

CREATORMengisti Berihu Girmay, Felix Möhrle
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE29 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2506-11665

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Perspectives on Explanation Formats From Two Stakeholder Groups in Germany: Software Providers and Dairy Farmers Mengisti Berihu Girmay, Felix Möhrle Abstract This paper examines the views of software providers in the German dairy industry with regard to dairy farmers’ needs for explanation of digital decision support systems. The study is based on mastitis detection in dairy cows using a hypothetical herd management system. We designed four exemplary explanation formats for mastitis assessments with different types of presentation (textual, rule-based, herd comparison, and time series). In our previous study, 14 dairy farmers in Germany had rated these formats in terms of comprehensibility and the trust they would have in a system providing each format. In this study, we repeat the survey with 13 software providers active in the German dairy industry. We ask them how well they think the formats would be received by farmers. We hypothesized that there may be discrepancies between the views of both groups that are worth investigating, partly to find reasons for the reluctance to adopt digital systems. A comparison of the feedback from both groups supports the hypothesis and calls for further investigation. The results show that software providers tend to make assumptions about farmers’ preferences that are not necessarily accurate. Our study, although not representative due to the small sample size, highlights the potential benefits of a thorough user requirements analysis (farmers’ needs) to improve software adaptation and user acceptance. 1 Introduction With the increasing advance of Artificial Intelligence (AI) , more and more applications are also emerging in agriculture. In arable farming, digitalization has enabled the concept of precision farming, where application rates of inputs (e.g. fertilizer) are precisely adjusted for different field zones. Traditionally, digital systems use rule-based programming logic (e.g. if-then-else). However, with its growing popularity, it is to be expected that AI will also be increasingly integrated into these practices. For example, satellite or drone images can be analyzed using image recognition methods to monitor plant growth and optimize cultivation — or to examine plants for diseases and pests in order to apply plant protection products at an early stage where necessary. AI also offers new possibilities for crop yield forecasts Berger et al. 2020 or predictions of the expected water consumption Linaza et al. 2021a . There is similar potential in livestock farming. Image-based or audio-based AI methods can be used to detect animal diseases (e.g. lameness or mastitis) or adverse animal behavior (e.g. tail biting). Further applications are likely to emerge in the future, and it is to be anticipated that AI will be a contributing factor to increasing the overall productivity of many farms Hoxhallari et al. 2022 ; Linaza et al. 2021b . While many believe that digitalization and, increasingly, AI hold great potential for the future, the adoption of digital solutions in practice is rather cautious, even though many technologies have been on the market for many years Dörr and Nachtmann 2022 . There are many studies that investigate possible influencing factors for the poor acceptance Pfeiffer et al. 2021 ; Mohr and Kühl 2021 ; Reichardt et al. 2009 . Farmers continually express dissatisfaction about the high complexity of systems as well as poor user-friendliness. Other factors include a lack of interoperability with other systems and concerns about data protection. A recent study with crop farmers has shown that farmers tend to stick with their current level of technology, suggesting a low likelihood that they will adopt new information-intensive technologies Gabriel and Gandorfer 2023 . Further research emphasizes that the explainability of digital systems plays a crucial role, as incomprehensible and non-transparent decision recommendations lead to a lack of trust among users Chazette et al. 2022 ; Shin 2021 . With the advent of AI in digital farming solutions, these adoption barriers are likely to rise even further. This is not least due to the black-box nature of many AI solutions, whose rationale is difficult for end users to understand (why is the system giving me this recommendation?). A number of recent studies have addressed the need for better explainability of digital systems. However, many researchers focus mainly on safety-critical use cases, e.g. in the automotive industry or healthcare Ahmad et al. 2018 ; Glomsrud et al. 2019 . With regard to the agricultural sector, we believe that not enough research is being done on the needs of farmers and, as a result, many digital solutions are not optimally adapted to their needs in practice. We see great potential for improvement particularly in decision support systems, which are designed to help farmers in their day-to-day decision-making. An example is the assessment of animal health and the recommendation of treatments. Without comprehensible reasoning, farmers may be reluctant to believe the system’s judgement. By providing a rationale (e.g. rise in body temperature or prolonged sitting) in addition to recommendations, the assessment can be made more comprehensible to farmers to increase their willingness to trust a system. Based on these observations, an entire branch of research has emerged over recent years, usually referred to as Explainable Artificial Intelligence (XAI) . The concept of XAI aims to make results generated by AI more comprehensible and transparent. In this respect, explainability can be considered a non-functional requirement that, if implemented appropriately, can contribute to better user experience. Recent focus has mostly been on the development of new XAI techniques, with little attention being paid to evaluation with stakeholders. Many authors, however, emphasize the need for stakeholder involvement to ensure meaningful benefits Vermeire et al. 2021 ; Vilone and Longo 2021 . In the agricultural ecosystem, farmers and software providers are two key stakeholder groups, each with different interactions and expectations of AI technologies Habiba et al. 2022 . Farmers are looking for directly actionable insights to make daily operational decisions. On the other hand, software providers may focus more on the functionality and innovation of algorithms, often at a level of complexity that is not easily understood by non-experts. This discrepancy in the perception of explainability can impair the benefits and acceptance for end users. In our work, we explore the notion of explainability of decision support systems using a small case study (online survey) from the German dairy farming sector, which we conducted over a period of 6 weeks. A hypothesis behind our study was that there could be discrepancies between the actual needs of farmers and the needs perceived by software providers, which could be one reason for the generally low acceptance of digital systems. To this end, we asked software providers for their views on the needs of dairy farmers regarding the explainability of digital systems. By software providers, we refer to employees of software development companies who are involved in the development of such systems. The study was designed with a relatively small scope and therefore cannot provide representative results. Rather, the intention was to provide initial findings and possibly a basis for future work. 2 Related Work There are different views in literature on what the term explainability entails and how it should be defined. We rely on the definition of Atf and Lewis 2023 , who refine explainability into two aspects, namely comprehensibility and trust. Comprehensibility refers to how well the user is

Read the complete text