Machine learning, being an important part of artificial intelligence, is increasingly discussed and rated in the media without explaining its functionality. This can lead to misconceptions of its real impact and range of application, a problem especially concerning young people. This contribution focuses on the theory-driven development and practical experience with an unplugged workshop concept, which is about a simple technique of machine learning, as a basis for possible teaching units for high school students. For this purpose, the focus of the workshop is an action-oriented method to simulate the classification of screws with two different lengths. Workshop participants can reconstruct linear classification by moving a classifier represented by a wooden strip according to defined rules after each insertion of training data on a pinboard. The aim is to examine whether and how the topic can be made understandable at school. Pre- and posttests are used to evaluate the impact of the workshop on the participants’ image of artificial intelligence and machine learning. The results of this research suggest that it is possible to reduce simple methods of machine learning for teaching this topic at school. Moreover, it seems that even a 90-min workshop can change the participants’ conceptions of machine learning and artificial intelligence to a more realistic appreciation of their impact.