AI-assisted exam marking where a human awards the mark
HiWi+ Exam Manager prepares the marking: answers grouped, notes referenced to your rubric, differences made visible. Assessment and decision remain entirely with the examiners.
HiWi+ Exam Manager is built and is currently being tested on real exams. We say so openly, because marking assistance is only worth anything if examiners trust it. Anyone who wants to be involved early can join as a pilot partner and see the module demonstrated on their own examination date.
Where marking assistance really saves time
Not in the assessment itself, that remains academic work. But in everything around it: sorting, comparing, justifying and documenting.
Group free-text answers
Similar solution paths are grouped so examiners assess comparable answers one after another instead of jumping across the pile.
Match against your rubric
Every note names the criterion from your own rubric that it refers to. Without a rubric there is no note.
Follow the working
For calculation questions it becomes visible at which step a result diverges, so partial marks can be awarded with a reason.
Prepare second marking
Differences between first and second assessment are listed so the discussion starts at the contested points.
Check equal treatment
Identical answers with different scores stand out. That is the first question asked in an appeal.
Recurring feedback
Common error patterns can be written once as feedback and then assigned, instead of being rewritten fifty times.
Work anonymously
Marking runs on candidate numbers. Names are neither needed nor shown for the marking process.
Record for the file
Every decision is recorded with time and person. That holds up when an assessment has to be justified later.
How a marking suggestion comes about
Four steps, and the fourth always belongs to the examiner.
Store the rubric
You provide your model solution and your marking rubric. The tool works exclusively against that specification.
Prepare the answers
Exam scripts are read in anonymised and grouped by question, so like is assessed with like.
Generate notes
For each answer a note is created with reference to the criterion. The note is a suggestion, not a score.
You decide
Accept, adjust or discard. Only your confirmation creates an entry, and that entry goes into the audit log.
The principles we do not depart from
Examination assessment is the area where a tool can do the most damage. That is why these points are anchored in how it is built, not stored as a setting.
What belongs to the module
Marking rubric
Criteria, scores and partial marks are stored once and reused across semesters. Changes remain traceable.
First and second marking
Both assessments run separately and are only compared afterwards. The second marker does not see the first assessment in advance.
Feedback to candidates
From the criteria awarded, a reasoned feedback text can be produced, which is reviewed before it is sent.
Audit log and export
The full marking history can be exported, for the file and in case of an inspection request.
What HiWi+ Exam Manager deliberately does not do
A marking tool that promises more than it can be held to is worth nothing in an appeal.
Karlsruhe University of Applied Sciences
HiWi+ is in use at Karlsruhe University of Applied Sciences and was developed together with the departments there, along real workflows rather than at a desk.
“Our main goal was to reduce administrative processes and workload within the faculty while at the same time offering students a better service, for example with the faculty chatbot they can ask questions about their studies around the clock. The HiWi+ modules were developed for our specific needs at the faculty, but they are certainly just as usable for many other educational institutions.”
What institutions ask before rolling it out
No. HiWi+ Exam Manager produces notes and suggestions with reference to your marking rubric. The score is entered exclusively by the examiner, and only that confirmation creates an entry. This is not a setting that can be switched off, it is how the module is built.
Assessing examination performance counts as a sensitive area of application. That is why the module is built as a pure assistance tool: it cannot decide anything with legal effect on an examination, every output is a reasoned suggestion, and human oversight cannot be bypassed. The full marking history is logged.
From your model solution and your marking rubric. It has no opinion of its own and no external yardstick. Without a rubric there are no notes.
Yes. For every suggestion it is visible which criterion and which passage in the answer it refers to. Suggestions without evidence are not displayed.
They are processed anonymised, on candidate numbers rather than names. No training takes place on your data. On request, processing runs in your own data centre or entirely locally via HiWi+ Local AI.
It is built and is in its validation phase. We are testing it on real exams together with examiners before it is used more widely. Anyone who wants to be involved early can join as a pilot partner.
Yes, and that is the sensible route. One subject, one rubric, one examination date. After that you can judge whether the effort for the pile of scripts and the quality of the notes match up.
Test it against your own rubric
We demonstrate the module on one of your exams and your own rubric. You then see on real answers how robust the notes are, before anything is decided.