Header

Search
Department of Informatics Human-centered Information Systems Engineering

Learning to Model with AI: How Learners Use LLM Support and React to Automated Feedback in BPMN Modelling

Conceptual modelling is a core competence in information systems, and BPMN is one of the most widely taught modeling languages. Learning it means acquiring both the syntax of the notation and the modelling judgement of what to represent — and feedback is what drives that learning. Automated, rule-based feedback (e.g., on syntactic correctness of a model) has long been used in modelling education because it scales to large tutorial groups.
With the availability of generative AI, students now routinely draft their models with LLM support, so a submitted model is increasingly a co-created artefact rather than the product of their own reasoning alone. If part of the modelling work is delegated to an AI assistant, students may invest less effort, feel less ownership of the result — and therefore process corrective feedback differently: possibly less defensively, but also less deeply. Whether AI support in modelling education accelerates or short-circuits learning  is an open question with direct consequences for how we design modelling courses.
In this thesis, you will extend an open-source BPMN modelling tool that offers optional LLM-based modelling support into an experimental platform that delivers automated rule-based feedback on the models students produce  while logging modelling behaviour in the background. You then run an experiment in an actual university course: students work on real modelling tasks either with or without AI support, receive automated feedback, and revise their models. The goal is to understand how students use AI when modelling, and whether AI-assisted students react to and act on feedback differently than students who modelled on their own.
 

What You Will Do

  • Review the literature across conceptual modelling education, automated feedback and intelligent tutoring in modelling, human–AI collaboration, and LLM support for business process modelling.
  • Build the experimental platform. Extend an open-source BPMN modelling tool with feedback and interaction logging.
  • Refine the experiment. Develop an experimental design (AI support vs. no AI support) embedded in a course tutorial, with self-report measures of effort, ownership, perceived feedback quality and affective reaction, complemented by behavioural log and model-revision data.
  • Run the study with students in a real tutorial setting, following ethical and data-protection requirements (ethics review, informed consent, debriefing)
  • Analyze the data. Relate condition to AI usage patterns, modeling quality, (affective) feedback reactions, and the extent and quality of subsequent revisions as well as learning progress.
  • Write up the findings, discussing contributions to research on human–AI collaboration and implications for the design of modelling tools and modelling courses.

Your Profile

  • Genuine interest in empirical, experimental research on human–AI collaboration and on how people learn with AI
  • Ability to implement a web-based prototype or adapt an existing open-source software (integrating an LLM API, setting up required logging, etc.)
  • Willingness to work with data in Python or R
  • Prior exposure to BPMN or conceptual modelling is helpful

The topic can be cut for a Bachelor's or a Master's thesis or adapted to fit the scope of a Master’s project). The exact scope of the prototype and thesis will be aligned during the proposal process before registration.
Please get in touch with a short email including your CV, a current transcript of records, and the planned start/finish date: hise@ifi.uzh.ch. This is a joint project of the HISE and ISIO research groups.