The continuous measurement of learning behavior or learning outcome by appropriate sensors is fundamental for the implementation of technology-based adaptive learning courses. An adaptive system needs such learning data to adapt the instruction to the needs of students. Depending on the learning objective, the sensors use information generated within tasks as basis for instructional adaptation, such as closed clear answers to learning tasks or more ambiguous open answers. In the second case, the use of self-evaluation concepts is one possibility. However, the automation and corresponding generation of learning data for adaptive systems is difficult because the answers are not clearly correct or incorrect. In a preliminary study we constructed a corresponding self-evaluation scaffold in the context of a Blended-Learning course in health psychology with 60 adult students. The goal of the study was to analyze if self-evaluation delivers valid data for sensors of an intended adaptive learning system, and what influences the accuracy of the self-evaluation. Therefore, we conducted an external evaluation of the students' answers. The correlation of the self-evaluation with the external evaluation was moderate (r=0.50; p < 0.001), and there was a large overlap of 65.9% between self- and external evaluation. The difference between self-evaluation and external evaluation can partly be explained by the task and subtask used, the difficulty of the task and the quality of the answers. In this respect, these results provide us with initial insights into what must be taken into account when designing corresponding sensors with an open response format. The analyses have also encouraged the development of a technology-based adaptive course on the basis of corresponding sensors. [For the complete proceedings, see ED600498.]