Abstract:

Choosing a suitable course of study is not an easy decision for prospective students. To make an informed choice, applicants should thoroughly research the various degree programs available. The information offered by different universities varies significantly. E.g., the enrollment processes differ between universities and even between faculties within the same institutions. Each study program has its own curriculum, which details the structure and processes of the individual programs. This wealth of information can be overwhelming for applicants. Additionally, the number of study programs offered in Germany almost doubled between 2007 and 2021.

AI-based technologies, such as chatbots, can help retrieve specific information on pre-prepared topics. They can be used regardless of location and time. Studies show that users often accept chatbots as digital assistants and generally rate them positively. At TH Köln/University of Applied Science, we developed an MIA – a chatbot for prospective students. MIA can provide information about the bachelor's and master's degree programs in mechanical engineering and offer advice on study suitability. We developed the chatbot through an interdisciplinary co-creation process involving academic staff and students and tested it in an initial study.

For the study, we developed a hybrid chatbot, which accesses the large language model GPT-4o using OpenAI's Assistant API, enabling it to generate language. For the basic structure of the chatbot, we used the Conversational AI Rasa. MIA's knowledge base was expanded using the RAG approach (Retrieval Augmentation Generation), incorporating processed study and examination regulations and additional information from websites. We prompted MIA to present the data in the role of a graduate of the respective study programs.

We conducted the exploratory study at a high school (upper secondary school) and used a questionnaire to measure the overall impression and acceptance of the chatbot and to gather general feedback. To measure participants' acceptance, the Unified Theory of Acceptance and Use of Technology (UTAUT) model was adapted for the use case. A total of 22 students participated in the study.

The results show that MIA is accepted and should be implemented for other study programs. The paper discusses the findings and outlines further possible steps. The chatbot should be expanded to include a stronger coaching aspect and be transferred to other contexts and tasks.

Keywords:

Chatbot, Conversational AI, RAG, LLM, study orientation, study assistant.