205 Artificial intelligence and the internet of things: responsible and sustainable intensification of livestock production.

This review compares and synergizes frameworks used in diverse disciplines to work toward responsible: artificial intelligence (AI); application of the Internet of Things (IoT); and sustainable intensification of livestock production. Although responsible AI is often called for as a key foundation of AI advancement, there is not yet a gold standard framework for responsible AI. Key principles applied across proposed frameworks include: transparency, fairness, accountability, and stakeholder integration. Many frameworks strongly advocate the need to apply these principles across the development, deployment, and use of AI tools. Responsible IoT, comparatively, is not widely discussed within the literature; however, several frameworks are proposed that contribute to responsible data management or network usage which are relevant within IoT systems. The FAIR data principles (findability, accessibility, interoperability, and reusability) are a widely used framework that is critical to data use within IoT systems; however, these principles do not extent to other elements of IoT systems. Trust-based frameworks, standardized integration methods, and interoperable platforms are additional examples of frameworks contributing to broader IoT responsibility. Frameworks for responsible sustainable intensification, similarly, are often diversely defined, with emphasis on implementation-oriented outcomes, and integrating a broad-basis of indicators and sustainable practises. Across these domain spaces, responsibility frameworks have emphasis on end-user integration, transparency and accuracy, with themes around fairness and equity. We apply these framework principles to evaluate emerging example uses of AI and IoT in sustainability domains for livestock production systems, including precision livestock farming (PLF), and measurement, modeling, reporting, and verification (MMRV) for environmental service markets. In the case of PLF, many systems are based on input from end-users; however, holistic incorporation of end-users into diverse aspects of system design is often limited. Transparency is often identified as a limitation of PLF technologies, sometimes due to communication issues, and sometimes due to challenges in technology development. Similarly, fairness and equity can be challenges because many PLF technologies focus on replacing human labor. Accuracy is often the major focus of research focused on PLF technologies, with other aspects presumed to be secondary considerations during development. Conversely, transparency is often the major focus of MMRV systems, with accuracy being a secondary consideration. Stakeholder integration and considerations for fairness and equity are treated differently through MMRV systems, with some having high integration, and others being developed without stakeholder consideration. These examples of early integration of AI and IoT technologies toward sustainability objectives can help inform future efforts focused on novel applications.