Clustering is a very common technique for finding similar records in a dataset, but classic clustering systems do not allow interaction or steering of the cluster generation process. Interactive clustering, the integration of clustering into visual analytics, on the other hand allows the analyst to interact during the cluster generation process. This can lead to more meaningful insights and can increase the trust and the transparency of the whole clustering process. Evaluations have already shown that these systems can produce clusters with lower error rates. However, there is little work on evaluating these systems with respect to insight generation. In this thesis, a literature study is conducted with the goal to provide guidelines for an insight-based evaluation of interactive clustering systems as well as interaction methods for interactive clustering systems. First, an overview of interactive clustering and different types of interaction is provided. Then, the term insight and three prerequisites (data, users, and tasks) for an evaluation are defined and described. Furthermore, different evaluation procedures are identified and analyzed due to their capability for the evaluation of insight. The results are then used to establish guidelines. They are useful for two reasons: It allows developers to obtain a better understanding of how well a particular system promotes insight generation and it can show which types of interaction between the analyst and the system can lead to high quality insights. In the end, the results are discussed and ideas for future work are provided.