AI Tutor in Project and Problem-based Learning

Project Description and Rationale

KI-TuProL is a joint research project of the Willy Brandt School of Public Policy at the University of Erfurt and the University of Applied Sciences Erfurt (FH Erfurt). The project focuses on building and testing an AI tutor that helps students with problem- and project-based coursework by asking them questions instead of giving them the answers.
 

AI chatbots are increasingly shaping the way students think and work. Steps where learning used to happen naturally, such as reading up on a topic, thinking through a problem or writing a first draft, can now be skipped in seconds. As a result, it is becoming harder to tell how much students have actually understood. This is especially noticeable in problem- and project-based courses, where students spend several weeks working on an open question from the real world. What interests us is how an AI tool would need to be designed so that it supports students' learning rather than doing the thinking for them.

Project Goals

The goal of this project is to develop an AI tutor that behaves like a good student tutor. It asks follow-up questions, points out things students may have missed and helps them organise their work, but the decisions and the actual work stay with the students. The tutor will also be set up for a specific course, so it already knows the course materials, learning goals and assignments. Students will not need to upload anything, and all data will be handled in line with the GDPR. 

Project Funding

KI-TuProL is funded by the eTeach-Netzwerk Thüringen.

Project Team

University of Erfurt: 

  • Lead: Dr. Hasnain Bokhari (Head of Digital Policy and AI)
  • Coordinator: Viddy Ranawijaya

University of Applied Sciences:

  • Lead: Prof. Rolf Kruse (Digital Media & Design / Applied Informatics)
  • Coordinator: Maximilian Röhr 

Our Approach

Three assumptions shape this project: 

  • First, students learn more when they are asked questions than when they are given answers. 

  • Second, a tutor that knows a particular course is more useful than a general-purpose chatbot, however powerful. 

  • Third, AI should support teachers and student tutors, not replace them. They remain responsible for the personal side of teaching, for grading and for setting limits.

Along the way, the project intends to find out how working with an AI tutor affects the way students experience their own learning, whether it helps them use AI more thoughtfully, and how far such a tutor can extend the support that teachers are able to offer.

Hasnain Bokhari
Dr Hasnain Bokhari
Head of digital policy and artificial intelligence
(Willy Brandt School of Public Policy)
C19 – research building "Weltbeziehungen" / C19.02.09
+49 361 737-4648
hasnain.bokhari@uni-erfurt.de
Profile page