Some tools at the University of Bremen are already working for you in the background in the Stud.IP learning platform:
Automatically generated AI feedback for free text answers in the assignment tool DoIT!
DoIT! now offers automatically generated AI feedback for free text answers from students. This feedback is only made available to teachers/tutors and can be used as a decision-making aid for their own feedback to students. An AI prompt with reference to the task and the answer is predefined as a placeholder for the tasks, i.e. it can be used or you can write your own prompts for the different tasks. The feedback is created at night after the processing deadline has expired, provided the students have agreed to this. This is based on an open source LLM (meta-llama-3.1-8b-instruct) from the GWDG Academic Cloud (https://chat-ai.academiccloud.de), i.e. it is used in compliance with data protection regulations.
Especially in large cohorts, the use of these feedback suggestions can lead to considerable time savings for teachers and tutors.
If you use the scoring system for the assignments, there is now also an AI-generated score suggestion the next day after the submission deadline.
From a didactic perspective, AI feedback can be used for various scenarios, e.g. for formative assessments. This can be the use as a pure tool for teachers/tutors to support the creation of feedback or to explore the possibilities of artificial intelligence as a teaching and learning object.
Didactic scenario: as a tool
AI-generated feedback is created for each task and used as a template for your own feedback. However, students can object to the use of the AI feedback when answering the questions.
Didactic scenario: as a teaching object
- Accept the AI feedback unchanged (mark accordingly) and have this output critiqued by students.
- Accept the AI feedback unchanged and mark it accordingly and also write your own feedback and mark it accordingly. This comparison of human and artificial responses can then be reflected on by the students.
- The students do not know whether the feedback was created by the AI or the teacher and analyze and evaluate the feedback (and they conduct a lecture hall survey, e.g. with Particify or the Stud.IP tool cliqr, and let the students vote on which feedback is human or artificial).
From a legal perspective, AI-generated feedback can support pre-correction in the context of summative assessments, but must not replace assessment and does not release students from the obligation of independent assessment (see page 3 of the report "Rechtliche Rahmenbedingungen und Anpassungsbedarfe zur Verstetigung des Einsatzes von KI-basierten Werkzeugen in der Lehre: Wissenschaftliches Rechtsgutachten im Auftrag der Universit?t Bremen" by Prof. Dr. Kirchner-Freis, Zitierlink: https://doi.org/10.26092/elib/3781).
If you would like to find out more about this or would like an introduction to DoIT! as an assignment tool in Stud.IP, simply contact the Stud.IP team at ZMML at infoprotect me ?!elearning.uni-bremenprotect me ?!.de
