In recent years, serious games have been established as an efficient medium in education and professional training. They have the capability to be effective tools to promote learning and encourage behavioral change, and they constitute a vital instrument for a variety of education and training scenarios. The combination of the serious gaming approach with role-playing is particularly promising, as authentic simulated environments provide mobile, safe and continuable settings for learners in which they can assume roles in specific contexts, explore new situations, and learn how to act and react without having to fear consequences in the real world.

A special challenge with this kind of games is shaping the pedagogical outcomes, as the effects generally depend on post-role-play reflection. Without feedback and reflection, the transfer to real world situations cannot be ensured. Computer-supported analyses can help to track and evaluate the learners’ performances, generate feedback, and provide structured recordings enriched with helpful features like integrated indexing, navigation instruments, search functions, and cross references between different media and data sources.

This thesis focuses on serious role-playing games for the professional training of social skills featuring intelligent support. Besides presenting the broad theoretical foundations, contextual background, and existing research in this field, it proposes a novel conceptual and technical framework for the design and implementation of such games and presents different case studies and evaluation approaches.

From the design perspective, the framework is characterized by chat-like interaction with scripted chatbots in a dialogic setting, a separation of immersion and reflection phases that is considered to be conducive for learning, and computer-generated adaptive feedback based on individual analyses. From the technical point of view, the framework is based on three main components: AI-controlled chatbots that adapt to the learners’ behavior, a multi-agent blackboard system to keep components independent and support performance optimization using parallel processing, and intelligent support for automated performance analyses and feedback generation.

This thesis presents two main case studies based on the framework. They showcase two different application fields for serious role-playing games for social skills training: conflict management (multi-user environment) and customer complaint management (single-user environment). Both training scenarios have been evaluated in mixed-method studies, combining different qualitative and quantitative methods of analysis to investigate how learners perceive the behavior of the chatbots and whether the training scenarios qualify as real training situations. Furthermore, the relation between the convergence of visual foci of attention and cooperation quality is analyzed in the collaborative conflict management scenario. The mixed-method approach including both, subjective and objective measures, allows for a complete and synergetic utilization of data and creates a solid foundation for drawing conclusions regarding the research objectives.