Digital Coaching

Digital Coaching

Digital Coach: Enhancing User Experience with Transparent AI Reasoning

Imagine interacting with an AI coach who doesn’t just give you answers but shows you exactly how it thinks and why it offers certain suggestions. The Digital Coach project explores precisely this idea, diving deep into how openly sharing an AI’s “thought processes”—known as intermediate reasoning—impacts user experiences, particularly in coaching scenarios aimed at personal growth and educational goals.

Why is this important?

Traditional AI systems often act like black boxes, delivering solutions without revealing their internal logic. This lack of transparency can lead users to distrust the AI, especially when unexpected outcomes occur. Our research shows that AI becomes significantly more effective when it openly displays its reasoning in a way that’s easy for users to understand, creating better engagement, trust, and outcomes.

What did we discover?

Through controlled experiments and real-world testing with students, we discovered that framing the AI’s internal reasoning clearly but humbly—using expressions like “I believe…” rather than assertive statements like “I know…"—dramatically improves user satisfaction. This modest framing sets realistic expectations, reduces disappointment from inevitable minor mistakes, and strengthens the relationship between the AI coach and the user, known as the “working alliance.”

Practical outcomes

  • Higher Trust and Engagement: Users who saw the AI’s intermediate reasoning, especially in a non-factive (“I think…”) manner, reported feeling their expectations were met more effectively.
  • Better Coaching Results: Students using this kind of digital coach showed increased satisfaction and reported better progress towards their personal educational goals.
  • Real Academic Impact: Regular interactions with transparently reasoning AI coaches correlated with higher academic performance.

What does this mean for you?

For anyone looking to integrate AI into coaching, education, or customer support, the Digital Coach project’s findings offer a clear roadmap. By carefully managing how AI systems express their thought processes, you can enhance user trust, satisfaction, and ultimately achieve better outcomes.

This research pushes the boundaries of how we interact with AI, emphasizing transparency and humility to create smarter, more empathetic, and effective digital companions.

Roman Rietsche
Roman Rietsche
Research Professor for Information Systems and AI

My research interests include AI in Education, Human-centered Re- and Upskilling Digital Interventions.