Refine
Document Type
- Conference Proceeding (5)
- Article (2)
- Part of a Book (1)
- Other (1)
Is part of the Bibliography
- yes (9) (remove)
Keywords
Institute
- Fakultät IuI (6)
- Fakultät AuL (2)
- LearningCenter (2)
In der Schriftenreihe „Voneinander Lehren lernen“ publiziert das LearningCenter der Hochschule Osnabrück anwendungsbezogene Beiträge zur Qualitätsentwicklung in Studium und Lehre. Die Schriftenreihe ist an das Konzept des „Scholarship of Teaching and Learning“ (SoTL) angelehnt. Demnach soll sie insbesondere den Fachlehrenden verschiedener Studiengänge als Plattform dienen, um ihre eigenen Erfahrungen, Ideen und Konzepte zur Lehr- und Studiengangentwicklung systematisch zu reflektieren und entsprechende Erkenntnisse für andere nutzbar zu machen. Ziel ist es, den Diskurs über hochschuldidaktische Themen in die Fächer zu tragen und so die Qualität der Lehr-Lernprozesse in den Studiengängen zu fördern. In diesem fünften Band der Schriftenreihe werden Projekte der Hochschule Osnabrück beschrieben, deren Umsetzung durch verschiedene Förderlinien oder durch Studienqualitätsmittel unterstützt wurde. Die Textbeiträge sind sowohl inhaltlich als auch didaktisch-methodisch sehr vielschichtig. Eine Gemeinsamkeit liegt jedoch darin, dass sie jeweils eine konstruktive hochschuldidaktische Reaktion auf zukunftsbezogene Trends und daraus resultierende Kompetenz-Anforderungen an Hochschulabsolvent*innen widerspiegeln. Der Terminologie des Zukunftsinstituts folgend sind es primär die Megatrends „Konnektivität“, „New Work“, „Gesundheit“, „Wissenskultur“ und „Globalisierung“, die in den Beiträgen implizit oder explizit thematisiert werden.
Compliance of agricultural AI systems : app-based legal verification throughout the development
(2024)
Significant advances in artificial intelligence (AI) have been achieved; however, practical implementation in agriculture remains limited. Compliance with emerging regulations, such as the EU AI Act and GDPR, is now vital, even for non-critical AI systems. Developers need tools to assess legal compliance, which is complex, often requiring full legal advice. To address this issue, we are developing a support app that simplifies the legal aspects of AI system development, covering the entire lifecycle, from conception to distribution. The current app, which covers the key legal area of copyright and will soon include GDPR and the AI Act, aims to bridge the gap between AI research and agriculture. An evaluation of our app by experts from both the legal and the IT domains shows that the app assists the developers so that they make legally correct statements. Consequently, it promotes legal compliance and awareness among developers, contributing to the seamless integration of AI into agriculture. The need for compliant AI systems in various industries, including agriculture, will only increase as regulations evolve.
Artificial intelligence (AI) promises transformative impacts on society, industry, and agriculture, while being heavily reliant on diverse, quality data. The resource-intensive "data
problem" has initialized a shift to synthetic data. One downside of synthetic data is known as the "reality gap", a lack of realism. Hybrid data, combining synthetic and real data, addresses this. The paper examines terminological inconsistencies and proposes a unified taxonomy for real, synthetic, augmented, and hybrid data. It aims to enhance AI training datasets in smart agriculture, addressing the challenges in the agricultural data landscape. Utilizing hybrid data in AI models offers improved prediction performance and adaptability.
Anwendungen wie ChatGPT oder WOMBO Dream machen es leicht, Studierende ohne Programmierkenntnisse für die Anwendung von künstlicher Intelligenz (KI) zu begeistern. Deshalb sind angesichts der zunehmenden Bedeutung von KI in allen Disziplinen innovative Strategien erforderlich, um Studierende ohne Programmierkenntnisse so auszubilden, dass die Anwendung von KI als Future Skill in die Studienmodule integriert werden kann. In diesem Artikel wird ein didaktisches Planungsraster für angewandte KI vorgestellt. Es orientiert sich am Prozess der KI-Programmierung (KI-Anwendungspipeline) und verknüpft KI-Konzepte mit studienrelevanten Themen. Diese Verknüpfung eröffnet einen neuen Lösungsraum und fördert das Interesse und das Verständnis für die Potenziale und Risiken von KI bei den Studierenden. Anhand einer Beispielvorlesungsreihe für Studierende der Energiewirtschaft wird gezeigt, wie KI nahtlos in den Unterricht integriert werden kann. Dafür wird das Planungsraster für angewandte KI an die Fachvorlesung angepasst. Dadurch können die Studierenden eine fachspezifische Aufgabenstellung Schritt für Schritt mit der KI-Anwendungspipeline lösen. So zeigt die Anwendung des didaktischen Planungsrasters für angewandte KI die praktische Umsetzung der theoretischen Konzepte der KI. Darüber hinaus wird eine Checkliste vorgestellt, anhand derer beurteilt werden kann, ob KI in der entsprechenden Vorlesung eingesetzt werden kann. KI als Future Skill muss von den Studierenden anhand von Anwendungsfällen erlernt werden, die für das Studienfach relevant sind. Aus diesem Grund sollte sich die KI-Ausbildung nahtlos in verschiedene Curricula einfügen lassen, auch wenn die Studierenden aufgrund ihres Studienfachs keinen Programmierhintergrund haben.
Artificial intelligence (AI) and human-machine interaction (HMI) are two keywords that usually do not fit embedded applications. Within the steps needed before applying AI to solve a specific task, HMI is usually missing during the AI architecture design and the training of an AI model. The human-in-the-loop concept is prevalent in all other steps of developing AI, from data analysis via data selection and cleaning to performance evaluation. During AI architecture design, HMI can immediately highlight unproductive layers of the architecture so that lightweight network architecture for embedded applications can be created easily. We show that by using this HMI, users can instantly distinguish which AI architecture should be trained and evaluated first since a high accuracy on the task could be expected. This approach reduces the resources needed for AI development by avoiding training and evaluating AI architectures with unproductive layers and leads to lightweight AI architectures. These resulting lightweight AI architectures will enable HMI while running the AI on an edge device. By enabling HMI during an AI uses inference, we will introduce the AI-in-the-loop concept that combines AI's and humans' strengths. In our AI-in-the-loop approach, the AI remains the working horse and primarily solves the task. If the AI is unsure whether its inference solves the task correctly, it asks the user to use an appropriate HMI. Consequently, AI will become available in many applications soon since HMI will make AI more reliable and explainable.
In modern times, closed-loop control systems (CLCSs) play a prominent role in a wide application range, from production machinery via automated vehicles to robots. CLCSs actively manipulate the actual values of a process to match predetermined setpoints, typically in real time and with remarkable precision. However, the development, modeling, tuning, and optimization of CLCSs barely exploit the potential of artificial intelligence (AI). This paper explores novel opportunities and research directions in CLCS engineering, presenting potential designs and methodologies incorporating AI. Combining these opportunities and directions makes it evident that employing AI in developing and implementing CLCSs is indeed feasible. Integrating AI into CLCS development or AI directly within CLCSs can lead to a significant improvement in stakeholder confidence. Integrating AI in CLCSs raises the question: How can AI in CLCSs be trusted so that its promising capabilities can be used safely? One does not trust AI in CLCSs due to its unknowable nature caused by its extensive set of parameters that defy complete testing. Consequently, developers working on AI-based CLCSs must be able to rate the impact of the trainable parameters on the system accurately. By following this path, this paper highlights two key aspects as essential research directions towards safe AI-based CLCSs: (I) the identification and elimination of unproductive layers in artificial neural networks (ANNs) for reducing the number of trainable parameters without influencing the overall outcome, and (II) the utilization of the solution space of an ANN to define the safety-critical scenarios of an AI-based CLCS.
While developing traffic-based cognitive enhancement technology (CET), such as bike accident prevention systems, it can be challenging to test and evaluate them properly. After all, the real-world scenario could endanger the subjects’ health and safety. Therefore, a simulator is needed, preferably one that is realistic yet low cost. This paper introduces a way to use the video game Grand Theft Auto V (GTA V) and its sophisticated traffic system as a base to create such a simulator, allowing for the safe and realistic testing of dangerous traffic situations involving cyclists, cars, and trucks. The open world of GTA V, which can be explored on foot and via various vehicles, serves as an immersive stand-in for the real world. Custom modification scripts of the game give the researchers control over the experiment scenario and the output data to be evaluated. An off-the-shelf bicycle equipped with three sensors serves as a realistic input device for the subject’s movement direction and speed. The simulator was used to test two early-stage CET concepts enabling cyclists to sense dangerous traffic situations, such as trucks approaching from behind the cyclist. Thus, this paper also presents the user evaluation of the cycling simulator and the CET used by the subjects to sense dangerous traffic situations. With the knowledge of the first iteration of the user-centered design (UCD) process, this paper concludes by naming improvements for the cycling simulator and discussing further research directions for CET that enable users to sense dangerous situations better.
Studies on nutrition have historically concentrated on food-shortages and over-nutrition. The physiological states of feeling hungry or being satiated and its dynamics in food choices, dietary patterns, and nutritional behavior, have not been the focus of many studies. Currently, visual analytic using easy-to-use tooling offers applicability in a wide-range of disciplines. In this interdisciplinary pilot-study we tested a novel visual analytic software to assess dietary patterns and food choices for greater understanding of nutritional behavior when hungry and when satiated. We developed software toolchain and tested the hypotheses that there is no difference between visual search patterns of dishes 1) when hungry and when satiated and 2) in being vegetarian and non-vegetarian. Results indicate that food choices can be deviant from dietary patterns but correlate slightly with dish-gazing. Further, scene perception probably could vary between being hungry and satiated. Understanding t he complicated relationship between scene perception and nutritional behavioral patterns and scaling up this pilot-study to a full-study using our introduced software approaches is indispensable.