COFFEE - Corrective Formative Feedback

Deutsch

Photo: CATALPA

COFFEE is an application that helps teachers provide feedback on free-text assignments to students. Teachers can enter their own assignments, along with the relevant assessment criteria and learning context, to efficiently create personalised feedback.

Using the COFFEE feedback application, students can receive specific feedback on their submitted solution at the touch of a button thanks to generative AI. Feedback can be generated for any number of questions and is not limited by course size, as it is automatically produced for each solution submitted. To get started quickly, typical criteria and suitable prompts are provided in the accompanying materials and can be easily customised. You can select Large Language Models (LLMs) to suit each criterion, allowing you to use the most appropriate model for the subject matter.
Screencasts provide an overview of the application:

Screencast students

Screencast teachers

Using COFFEE, teachers can provide feedback to students on solutions to free-text assignments. This feedback is based on didactic criteria specified by the teachers themselves. In simple terms, COFFEE works as follows: Teachers enter their tasks and assessment criteria into the application, which sends them to a generative AI model via a prompt. The model then sends out solution-specific feedback. This enables students to swiftly identify the strengths and weaknesses of their solutions in relation to the course's learning objectives. COFFEE was developed through a value-sensitive design process and released as an open-source solution for universities. Once installed and the user groups have received AI training, COFFEE can be used directly. All the necessary information and accompanying materials can be found in the Media section. Framework conditions such as data protection and the EU AI Regulation were considered during the development process.

    • Wannemacher, K., Bosse, E., Lübcke, M., & Kaemena, M. (2025). Wie KI Studiumund Lehre verändert. Anwendungsfelder, Use-Cases und Gelingensbedingungen (Arbeitspapier (Geschäftsstelle Hochschulforum Digitalisierung) No. Nr. 87; S. 74ff). Hochschulforum Digitalisierung. https://hochschulforumdigitalisierung.de/publikationen/arbeitspapier-nr-87-wie-kistudium-und-lehre-veraendert/
    • Karolyi, H., van Rijn, L., Hanses, M., de Witt, C. (2025). Ein wertesensibles Design für formatives Feedback mit Trusted Learning Analytics und KI, in: Queckenberg, R., Leschke, J., Persike, M. (Eds.), Learning Analytics, Artificial Intelligence und Data Mining in der Hochschulbildung, Beiträge zur Learning Aid 2024. transkript Verlag, Bielefeld, pp. 101–111. ISBN: 978-3-8376-7583-2
    • Hanses, M., Karolyi, H., Wöhrle, J., De Witt, C. (2025). In-House GenAI for Corrective Formative Feedback in Higher Education, in: Tammets, K., Sosnovsky, S., Ferreira Mello, R., Pishtari, G., Nazaretsky, T. (Eds.), Two Decades of TEL. From Lessons Learnt to Challenges Ahead, Lecture Notes in Computer Science. Springer Nature Switzerland, Cham, pp. 319–324. ps://doi.org/10.1007/978-3-032-03873-9_43
    • Zesch, T., Hanses, M., Seidel, N., Aggarwal, P., Veiel, D., & de Witt, C. (2024). Flexible LLM Experimental Infrastructure (Flexi) – Enabling Experimentation and Innovation in Higher Education Through Access to Open LLMs. In 2024 21st International Conference on Information Technology Based Higher Education and Training (ITHET) (pp. 1–8). IEEE.https://doi.org/10.1109/ITHET61869.2024.10837635
    • Wöhrle, J., Hanses, M. Karolyi, H., van Rijn, L., de Witt, C. (submitted). University Students’ Adoption, Feedback Perceptions, and Learning with a Generative-AI Tool: A Pre–Post Study of COFFEE In. European Association of Distance Teaching Universities. Rethinking Higher Education: Engage, Adapt, Include. Proceedings of the Innovating Higher Education Conference 2025.
    • Karolyi, H. et al. (accepted). COFFEE - KI in der Hochschulbildung nachhaltig nutzen. In Netzwerk Hochschulforschung Österreich (Hrsg.), Innovate. Adapt. Preserve: Navigating Change in Higher Education, Tagungsbad der 5. HoFo Konferenz Waxmann.

  • Code of the application COFFEE incl. 1 LLM (Phi:4):
    https://github.com/hansesm/COFFEE

CATALPA | 25.11.2025