AI-generated image of android cows grazing on a commons in Switzerland.
commons noun

A community-governed resource or space managed for shared benefit. Access and responsibilities are defined by clear rules—historically meadows, forests, and water systems; today also open-source code, data, and knowledge.

Translations

Deutsch (CH)
Allmend (Allmende) — gemeinschaftlich genutzte Flur/Weide.
Français
les communs (biens communs) — ressources gouvernées en commun.
Italiano
beni comuni — risorse gestite dalla comunità.
Rumantsch Grischun
bains cumüns — resursas gestiunadas communablamain.

Ethel Commons builds on the Ethel project (/ˈɛθəɫ/), which has supported teaching and learning at multiple Swiss universities for four semesters. Many institutions find it challenging to establish and maintain their own AI infrastructure. By working together, we can operate secure, scalable, and affordable AI for Education as shared infrastructure.

Purpose-Built

Ethel is built by educators for educators:

  • Course chatbots that use course materials as references
  • Formative assessment:
    • Homework feedback on handwritten work, PDFs, Word, and Jupyter Notebooks
    • On-demand interactive practice problems (proof of concept)
    • Quick-poll lecture feedback
  • Summative assessment: AI-assisted grading of handwritten exams (prototype)

Open Source

Ethel is released under the GNU GPL. Anyone can inspect the code, and any institution can participate in its ongoing development.

The current production version runs on Docker Swarm. It is not exactly beautiful, but it has been working for four semesters for thousands of students.

The next version is more modular: Kubernetes microservices orchestrated with LangGraph flows.

Compliant

Data stays in European data centers and is not used for model training. Operations are aligned with the EU AI Act. Privacy and governance are first-class concerns.

We plan to default to state-of-the-art OpenAI models via Azure in a Swedish data center; other models can be connected if scalable inference is available via standard APIs.

Research-Based

Developed and studied across four semesters. Selected publications:

Additional preprints:

No miracles, no gimmicks

We believe that humans learn best from and with other humans.

We aim to continue building solutions that support research-based educational methods, making them more scalable to larger numbers of learners.

AI should be a tool that works in the background, while humans remain front and center.

Commodity

We believe AI will become a campus commodity—like electricity, water, and the internet.

Based on average usage over four semesters, our target institutional pricing is:

  • CHF 1.25 per student per month
  • CHF 0.75 per exam

Let's see if we can provide equitable access for all learners through institutional adoption.