OSATI is a collaborative project between the Tübingen Center for Digital Education, the European Laboratory for Learning and Intelligent Systems (ELLIS) and the Tübingen AI Center.

Existing generative AI systems offer significant potential for supporting students by providing extensive answers to almost any questions. However, numerous studies have shown that students learn particularly effectively through interactions and targeted support measures, as these methods enhance their motivation and understanding of problem-solving. These studies have also demonstrated that human teachers are the best tutors, partly due to their elaborate feedback and support measures. However, not all learning phases involve a human teacher's presence (e.g., during homework, exam preparation, or self-study periods).

The goal of the OSATI project is, therefore, to develop a digital tutor that can provide the best possible and adaptive support to students during these self-study phases by interacting with them and guiding them through self-learning processes. This will be achieved through the training of artificial intelligence that acts as closely as possible to human tutors and can systematically assist in the learning process. The data foundation for training the AI-supported tutoring system will be real interactions between human tutors and students. The consideration of media ethics and media didactic aspects is central to this endeavor. The overarching aim of the project is to complement the generation of datasets, tools, and models for scalable open-source AI tutoring systems