Images play a central role in the context of digital products. They create identification and convey emotions. The widespread use of digital products on different devices and platforms results in special requirements for the provision of this image content. Due to different screen resolutions and screen formats, image media often have to be provided in different versions. For this purpose, it is often manually cut to the required formats and scaled according to the various requirements. These activities tie up a large part of the working time of content managers. AI-based solutions already exist on the market that can support and partially take over manual image cropping. It is therefore obvious to use these technologies and integrate them into the application in the form of intelligent system components. For the selection of the appropriate depth of integration as well as the choice of a suitable degree of interaction, the underlying problem domain is investigated by the seven-step process of Mackeprang et al. [1]. This enables a systematic analysis and supports the identification of possible task assignments between humans and computers. A task assignment corresponds to a list of the necessary substeps of a process, as well as an assignment of these substeps to the executing interaction partners. In the course of this work, three of these task assignments are concretized and transferred into functional program code. An evaluation is performed using predefined performance metrics. By including human as well as algorithmic perspectives, a multidimensional evaluation is made possible. In addition to the processing time and the quality of the image cuts, the workload of the users will also be surveyed. Finally, the task assignment that is found to be most efficient during the evaluation is used to formulate a recommended course of action for the automation problem under investigation.

[1] M. Mackeprang, C. Müller-Birn, and M. Stauss. Discovering the sweet spot of human—computer configurations: A case study in information extraction. Proceedings of the ACM on Human-Computer Interaction, 3 (CSCW), 2019.