The MachineLearnAthons, represents a new teaching format based on ML challenges seeking to develop ML competencies among students with and without data science backgrounds using real-world use-cases as application grounds. ML challenges are at the core of our concept, because they lead to a high degree of motivation among students (Makridakis et al. 2022). Since ML pipelines can be neigh endlessly improved, focusing on the challenge can keep more experienced students engaged allowing for a successful deployment in heterogeneous teams.The project contributes to the “democratization” of ML. We convey the necessary competencies to students with and without programming skills to develop ethically sound and robust ML solutions. On the one hand we address students with backgrounds in Engineering, Business Administration, or Logistics with little to no experience with programming and statistics. On the other hand, we address Computer Science students with some knowledge of statistics, and ML but lacking the experience of working on real-world data sets in interdisciplinary teams.
Teams working on ML projects need to incorporate both (methodological) ML knowledge and domain expertise. Due to the accelerating pace of digitalization in all aspects of life, data literacy, and ML competencies specifically are increasingly relevant key competencies. Given the complex technical and societal context within which ML models are to be deployed, ML solutions need to be not only highly performant but also trustworthy, safe, fair and register a high degree of acceptance. ML methods are a difficult subject for students with little to no prior programming and ML method experience. Conversely, it can be challenging for ML engineers to explore a completely new application domain. However, this domain knowledge is essential to creating useful ML solutions. ML challenges make it possible to combine the two types of expertise into inter-disciplinary teams.
The project contributes to the “democratization” of ML. We convey the necessary competencies to students with and without programming skills to develop ethically sound and robust ML solutions. On the one hand we address students with backgrounds in Engineering, Business Administration, or Logistics with little to no experience with programming and statistics. On the other hand, we address Computer Science students with some knowledge of statistics, and ML but lacking the experience of working on real-world data sets in interdisciplinary teams.
The objective of our project is the development and implementation of the new, challenge-based teaching format we dub “MachineLearnAthon”. The teaching concept is subject to the following principles:
- ADVANCING AWARENESS AND SKILLS IN MACHINE LEARNING: The format aims to raise awareness about the strengths of machine learning (ML) and develop competencies in understanding, preparing, developing, evaluating, and implementing ML solutions.
- PROVIDING CLARITY ON ML RISKS AND LIMITATIONS: The format emphasizes understanding the ethical and technical risks and limitations of ML, such as issues related to ethics (e.g., discrimination, bias) and technical aspects (e.g., extrapolation, interpretability, robustness).
- ENHANCING DATA LITERACY: The format focuses on developing data literacy among students, i.e. enabling them to effectively handle data in a systematic manner, even if they are not data scientists themselves (GI 2018).
- ENCOURAGING INTERDISCIPLINARY COLLABORATION AND INTERNATIONAL COOPERATION: The format aims to equip students with the necessary skills for collaborative work in interdisciplinary and international teams. It also ensures that students without prior ML knowledge are not excluded and fosters appreciation for domain expertise.
- PROMOTING ACTION-ORIENTED LEARNING: MachineLearnAthons are designed to be engaging and motivating, utilizing relevant data sets and use-cases within students’ domains. Following project-based learning principles (Düdder 2021a), the format encourages self-learning and active participation.
To achieve these goals, real-world use-cases are essential. As such, we pursue a tight cooperation with industry partners. This cooperation benefits all involved parties: The academic consortium can improve their teaching and research activity using concrete, motivating examples to showcase ML methods. Conversely, the industrial partners can gain additional insights from teachers potentially leading to improved solutions. Additionally, companies gain access to an additional talent pool in the form of MachineLearnAthon participants. Last but not least students gain perspective with respect to ML application “in the wild”. This means that, on the one hand, rather than working with ML ready datasets, i.e. clean, well structured, well documented etc., they have to deal with problems they would normally be confronted with within the context of a data science job. On the other hand, they need to hone their communication skills and be able to present their results to non-technical personnel.