1 Introdution
If feedback is considered as a criterion for automated support in learning, the device presented by Pressey in 1923 was the first teaching machine (Ludy 88). Skinner picked up Pressey’s design as well as the foundation in the theory of Thorndike. Based on Skinners concept machines and systems for programmed instruction (PI) were developed. Skinners concept is well known as the basis for behavioural learning theories. Extended computational power and general problem solving theories lead to the idea of intelligent tutoring systems (ITS) and adaptive learning environments (ALE). The idea was first based on the concept for the General Problem Solver (GPS), where the knowledge of problems and strategies to solve problems were separated. When the GPS failed for any relevant problem, the concept was replaced by expert systems. The core architecture of the DENDRAL expert system (knowledge base, explanations system, inference engine) became the starting point for SCHOLAR (Burton 1989), which was build as a semantic network and based on the architecture of expert systems. These concepts are closely related to cognitive learning theories. Despite the effort invested in ITS, there are hardly actually working systems available or real world applications reported. ITS seem to have failed due to the high effort necessary to develop such systems and the lack of theoretical foundations (Schulmeister 2007). In the last years, the successful application of recommender systems in marketing led to the idea of transferring those systems in the didactical field in the form of educational recommender systems (ERS) (Duwal 2011). This often takes place in the context of informal learning processes. The concepts seem to be related to constructivistic learning theories, while explicit references are rare. While most of the suggested ERS are in the early stages of development, the expectations are high. At least, these expectations appear to be similar to the systems discussed before.