Abstract:

Universities have been testing the use of chatbots for a long time, especially script chatbots that can provide an answer to a question within the limits of a prepared dialogue script. With the successes in machine learning, chatbots with artificial intelligence (AI) are a new hope for service-oriented and yet legally and data protection compliant use at universities. Many users at universities are currently testing the AI language model ChatGPT.

However, it is questionable how ChatGPT as a language model can be used sensibly in the teaching, administration, and research of a university. Current user experiences almost exclusively relate to the use of the ChatGPT dialogue, which involves a freely formulated inquiry, with follow-up questions and answers coming from an open dataset that the AI has been trained on.

For university purposes, this type of use can be disadvantageous, for example, because the university cannot control the communication request specifically, incorrect information from the used standard data set can be outputted, or disadvantages for students can arise who do not want to be pushed to register with ChatGPT in a data protection-compliant manner. However, the interface to AI language models allows universities standardized and imperceptible access to the possibilities of text AI, even without a student's registration.

This article focuses in particular on the possibility of integrating ChatGPT into a campus management system, specifically in TraiNex (www.trainex24.de). By using the API, the free inquiry prompt can become a mixed inquiry from known data (e.g., the field of study or the semester), requested data (e.g., research interests), and established data such as the university's framework conditions (e.g., prescribed structure). These data can be generated in the everyday usage process at an appropriate location in the campus management system and combined as a prompt. The prompt is imperceptibly passed through the API to the AI, which processes the inquiry in the background and returns the answer. The answer is analyzed by the campus management system, if necessary post-processed, and presented to the user. The user themselves does not see the prompt or know that an AI was used in the background for the generation of the response.

The article first presents an empirical study in which 5000 students were addressed by a chatbot. It shows that about 50% of all questions could have led to legal problems for the university if answered incorrectly. Based on this, a prototypical project is shown where instead of using a chat, students communicate indirectly with the AI via the API. It is shown how a professional prompt can be generated through the campus management system. Practical implementations such as "question to a course material," "subject-specific motivation of the student," "question to the examination regulations," or "help in finding a creative title" are presented (http://titelfinder.trainex.rocks?langu=2).

Keywords:

Chatbot, Campus management system, university, chatgpt.