Abstract:

Students face a variety of personal challenges during their studies, learn to deal with changes in their lives and master unforeseen tasks. To support and accompany students through such and similar challenges, universities are increasingly developing coaching services. They stimulate students' self-reflection processes to make possible actions visible and promote solution-oriented thinking.

To make coaching offerings scalable, digital technologies such as chatbots are increasingly being used in higher education. Coaching chatbots have enormous potential here. They can serve as a low-threshold offering through which students can deal with sensitive topics without judgment or shyness.

At our university, we are currently developing a coaching chatbot for students on the topic of exam anxiety. In the conception, we are looking in particular at how we can design the chatbot in such a way that a human-machine relationship can be established. Especially in coaching and with sensitive topics like exam anxiety, building a bond or rapport is a key success factor. In the accompanying research, we are therefore investigating which relationship-building factors can be effective in chatbot coaching.

One factor in building a human-machine relationship is the concept of self-disclosure. Self-disclosure in the case of chatbot coaching means that the chatbot discloses something about itself, which leads to the counterpart, i.e., the user, also disclosing something about itself. The relevant literature and numerous studies in human-machine interaction demonstrate the influence of disclosure behavior on constructs such as trust, rapport, or the establishment of a working alliance.

In the planned paper, we present the results of a study comparing different disclosure behaviors of a coaching chatbot and measuring their influence on relationship-building concepts such as working alliance and rapport. For this, we develop three coaching chatbots that coach on the topic of exam anxiety. In doing so, we only vary the disclosure behavior: The chatbot either shows self-disclsoure, i.e., it reveals something about itself and its exam anxiety. Or it shows information-disclosure, i.e., it reveals something about itself as an expert who has experience with students suffering from exam anxiety. In the third experimental condition, the chatbot does not exhibit any disclosure behavior. As the conversation progresses, we operationalize the disclosure behavior such that each coaching question asked by the chatbot is preceded by a disclosure statement to strategically encourage users to also disclose something about themselves.

In our study, we present our subjects with excerpts from an interaction with our coaching chatbot in which we varied the disclosure behavior. In a questionnaire, we ask them which chatbot with which disclosure behavior they were able to build a stronger relationship with. The research question we are interested in is: What is the impact of a coaching chatbot's disclosure behavior on relationship building, rapport and working alliance? To answer the research question, we hypothesize the following: Students perceive a stronger human-machine relationship when the chatbot shows self-disclosure than when it shows information-disclosure or no disclosure.

The goal of this work is to gain insights into how we need to design a coaching chatbot at universities to create a human-machine relationship, and for coaching to be effective.

Keywords:

Chatbot, Coaching, Self-Disclosure, Rapport, Working Alliance.