Announcement ยท
Introducing Feedbacker.
Feedbacker is free, open-source software that helps educators in higher education use AI to mark and moderate coursework against a rubric and brief. The AI suggests; you decide every mark, and approve every word a student receives.
Why I built it
Feedbacker began years ago as a small tool to help markers write meaningful, consistent feedback. This version started when I was asked to moderate a colleague's marking, and wanted a second reader that would help my judgement without replacing it. AI can now read a submission against a rubric and say why; the hard part is everything around that. Who decides the mark? What does the AI see? Can anyone check later what it suggested, and what was done with it?
Feedbacker is my answer to those questions. It is a way of working with AI on assessment, not a feedback generator.
What it does
- Marking a cohort, with feedback. Import a whole cohort from Turnitin or Canvas. Mark each submission against the rubric, with the AI's suggested levels before you mark, after you mark, or not at all. The AI drafts feedback from your own marks and comments; Feedbacker checks every piece against its mark, and you approve exactly what each student receives.
- Moderation. Mark a sample of a colleague's work yourself, openly or blind, with an independent AI reading alongside, then see where you, the original marker and the AI agree, and record your verdicts.
What it never does
It never gives a mark, and nothing reaches a student without your approval. It isn't a website you upload work to: it runs on your own computer. Before anything can be sent, it replaces the names, IDs and contact details in the text that its rules find. The rules can miss things, such as a name they don't know, and images aren't redacted at all, so you review each text and choose which images to send: your review is the safeguard. Nothing goes to the AI until you have approved it and confirmed the request. Every AI suggestion records which AI made it, its instructions and what it was sent.
Where it stands
I have used it for real: for a moderation, where it saved me time and helped my judgement rather than replacing it, and to mark and give feedback to a small cohort. Version 0.1.0 is out.
It is early. There is no installer yet, so setting it up takes a few technical steps. It works in Chrome or Edge, with Anthropic's Claude and your own API key, and with Turnitin and Canvas downloads of typed work. Marking in a team, and agreeing standards across markers, are still to come.
Find out more
If you mark or moderate, or support people who do, I would like to hear what you make of it.