Research & Innovation

Researching Trustworthy AI for Teaching and Learning

AI-based Content Co-Creation & AI Tutor Twin is a 27-month research and development project. We are exploring how generative AI can support educators and learners in accredited higher education while staying curriculum-aligned, overseen by people, and traceable.

Project period

Jul 2026 to Sep 2028

Funding program

ProFIT Brandenburg

Status

Research in progress

Kofinanziert von der Europäischen Union

Gefördert aus Mitteln der Europäischen Union und des Landes Brandenburg.

Funded with resources from the European Union and the State of Brandenburg through ProFIT Brandenburg, administered by Investitionsbank des Landes Brandenburg (ILB).

Funding details

The Challenge

AI Is Already Part of Studying. Universities Need to Be Able to Trust It.

General-purpose AI tools can produce content that is not grounded in approved sources. A chatbot that hands learners the answer can undermine the learning it is meant to support. For universities that award accredited degrees, both are real problems.

Grounding

Course content has to match the official module handbook and approved sources, not just sound plausible.

Learning

Support should help learners reason, not replace the reasoning that assessment is there to verify.

Accountability

Institutions need to see and explain how AI influenced teaching and learning.

What We Are Researching

Two Modules, One Question: How Can AI Support Learning Without Taking Control Away From People?

For educators

AI Content Co-Creation
Challenge Studio

An AI co-pilot designed to help educators turn their expertise and source material into course content and assessment questions that match the official curriculum.

Research question

How can AI help educators turn their expertise into curriculum-aligned courses, with oversight where it matters?

Expertise + sources

→

AI draft

→

Curriculum check

→

Educator oversight

Educators stay in charge, and sign off where it matters.

For learners

AI Tutor Twin
A Learning Companion

A personal learning companion designed to be available around the clock, adapt to each learner's progress, and guide with questions instead of handing out answers.

Research question

How can personalized AI support strengthen learning without undermining assessment?

Can you just give me the answer to task 3?

Let's get there together. Which concept from Module 2 do you think applies here?

Illustration of the research concept, not a product screenshot.

The Research Approach

What Makes the Approach Different

Instead of relying on a language model alone, the project pairs generative AI with a rule-checking layer. The same approach applies to both modules: it is designed to compare what the AI produces against official module handbooks and approved sources, route higher-risk output to an educator, and record each AI decision so it can be traced and reviewed.

Generate

AI drafts course content or answers a learner.

Verify

A rule-checking layer compares it with module handbooks and approved sources.

Escalate by Risk

Routine output goes through. Higher-risk output goes to an educator.

Trace

Each AI decision is designed to be logged so it can be reviewed later.

Why this matters now

Some uses of AI in education fall within the high-risk framework of the EU AI Act. That makes human oversight, traceability and alignment with approved sources design principles from the start of the project.

Design Principles

AI Supports. People Stay in Charge.

Human in the Loop

Educators oversee the system and step in where the stakes are higher. Assessment-relevant content always gets educator sign-off.

Curriculum-Aligned

Built around official module handbooks and approved sources.

Made for Accredited Higher Education

Designed around the quality standards of a state-recognized university.

Traceable by Design

AI decisions are meant to be recorded and reviewable.

Expected Results

What the Project Aims to Deliver by September 2028

Research prototypes for curriculum-aligned, AI-assisted content creation

An adaptive tutoring approach that guides learners rather than revealing answers

Mechanisms for personalization and for tracing AI decisions

Validation in real degree programs at Tomorrow University

The groundwork for offering both modules to other universities

Project Timeline

From Research Question to Validated Prototypes

Jul 2026 · We are here

Research Begins

Research and development of both modules is underway.

2027

First Research Prototypes

Testing the core content engine and the tutoring approach.

2028

Personalization, Traceability and Pilots

Further development and testing in real degree programs.

Sep 2028

Project Completion

Results published. Depending on outcomes, preparation for use at other universities.

Building on earlier research

ProFIT Berlin, 2022 to 2024

Our earlier ProFIT Berlin research project created the personalized learning engine that runs at Tomorrow University today. This project extends that work into content creation and tutoring.

Learn more about ProFIT Berlin

Funding