Hector AI + Education Future Fund

Empowering Informatics Teachers in AI through Professional Development and postgraduate Education

The Challenge 
In Baden-Württemberg, the subject of media education and computer science for grades 5-11 will be successively introduced at all general secondary schools from the 2025/26 school year. Courses in the Upper Secondary School are also planned in the long term and should enable a subject-specific and didactically sound Abitur in computer science. The computer science teachers currently active in BW cover the current demand to some extent, but not the broad scaling to a compulsory subject including Upper Secondary School. BW must therefore quickly and yet thoroughly qualify more computer science teachers and impart the urgently needed subject-specific and didactic competences in AI.
Our Solution
In cooperation with the Centre for School Quality and Teacher Training (ZSL) and the Ministry of Education, we will develop a modular tthat is conceptually based on the IMP contact study program. Training courses developed by us and certified by the ZSL will be offered quickly after the start of the project to reach as many teachers as possible.We enable computer science teachers in BW in all phases of their training and professional life to offer technically sound and didactically excellent computer science lessons for their students at both lower and upper secondary level.

AI-supported literacy development in kindergarten – A mobile app for authentic language support in early education.

The Challenge
Language support remains one of the central challenges in early childhood education. Despite substantial efforts by the federal states, accompanying studies on the effectiveness of language support measures consistently point to limited impact – with significant implications for the acquisition of written language: by the end of primary school, one in four children fails to meet the minimum level of reading comprehension required for continued educational success (IGLU 2021). While there are comparable measures to address these challenges, their implementation places considerable demands on educational staff and requires substantial personnel resources.

Our Solution
The AI-LIT project seeks to develop a mobile, AI-based application for dialogic reading, aiming to systematically support language interaction among preschool children during shared picture book reading. The project will implement dialogic reading (Whitehurst et al., 1998) using digital picture books in a scalable and fully automated format based on current AI technologies. Existing digital tools in this domain largely focus on supporting reading processes in primary education. In contrast, AI-LIT addresses language development at an earlier stage, aiming to support receptive and productive vocabulary, academic language skills, and narrative competence prior to school entry.

COMPASS, short for Comprehensive Open Math Platform with Adaptive Self-Regulation Support, aims to leverage AI to foster domain-specific learning while developing self-regulated learning strategies through an adaptive math learning platform supporting tens of thousands of pupils across Germany.

The Challenge
Self-regulated learning is crucial for developing lifelong learning skills, critical thinking, and academic resilience. Adaptive digital AI tools present both opportunities and challenges for developing self-regulation skills: they might undermine the students' ability to develop independent self-regulation by encouraging reliance on algorithmic guidance, or they could enhance these skills through personalized support. How much and what kind of adaptive support versus self-regulation do individual students need at different learning stages? 

Our Solution
An open math platform will systematically assess and optimize various self-regulation strategies, ensuring positive impacts on students' autonomous learning competencies and overall educational outcomes.
In detail: We build an adaptive mathematics learning platform starting from MatheBattle (mathebattle.de), an established platform in Baden-Württemberg with >10 million completed tasks yearly and nearly 100,000 active users. We redevelop MatheBattle as a modern, open-source React application based on the Growify Learning platform, migrate existing users, and integrate adaptive tutoring components. We first introduce an AI-based task editor for interactive learning content, and add features prompting learners to engage in self-regulated learning (e.g., goal setting, self-evaluation). We then add an evidence-based tutoring model engaging students in self-regulated learning strategies, and finally implement AI-driven knowledge tracing and adaptive learning paths aligned with US Common Core Standards (mapped to German curriculum).

Enhancing teaching quality with artificial intelligence: a feedback system leveraging AI to provide teachers with real-time, formative feedback on their teaching practices.

The Challenge
Even though the quality of teaching is a key determinant of effective learning, getting timely, reliable feedback about one's teaching quality remains a significant challenge. Traditional methods (student evaluations, teacher self-assessments, classroom observations etc.) have limitations in predictive validity, are time-intensive and costly and lack the actionable depth needed to foster meaningful, ongoing professional growth. This gap prevents teachers from reflecting on and adapting their daily work practices if they strive for educational excellence.
Our Solution
Real-time, formative feedback on teaching practices: We aim to develop an AI-based copilot application called Teaching Copilot that teachers can use on their smartphones to self-assess the quality of their teaching. Teachers can record classroom audio using their smartphones, and the system analyzes this data (audio and transcripts) to predict teaching quality aspects such as instructional organization, cognitive stimulation, or encouragement and warmth. Teaching Copilot will be a self-assessment tool empowering teachers to self-reflect. The system adheres to GDPR and EU AI Act regulations and privacy-preserving machine learning (ML) to ensure secure, ethical data handling. This innovation will enhance teaching effectiveness, enrich student learning experiences, and promote a culture of professional growth.
 

Voice-operated intelligent learning assistant for gifted children: An evidence-based open-source AI tutor for gifted elementary students.

The Challenge
Giftedness is regarded as a crucial incubator for driving innovation, productivity, and societal advancement. However, the recent PISA-findings revealed that gifted children are increasingly neglected: the share of adolescents who were able to translate their giftedness into high achievement has nearly halved in 10 years. Support for gifted students should begin early and be sustained continuously to nurture the curiosity and competence of the next generation of talented students. We see distinct potential in the use of AI, as it allows for seamless, adaptive, on-demand tutoring.

Our Solution
The goal of this project is to develop an evidence- and voice-based open-source AI tutor for gifted elementary students, starting with a tailor-made Hector Core Course and later extending towards general topic spaces. This AI tutor acts as a personal tutor for scientific questions and is designed to inspire discussions and offer age-appropriate scaffolds (e.g., prompts, hints, explanations) that guide children's curiosity-driven discovery learning, both inside and outside of school. To ensure success, we follow an agile, co-constructive, and practice-oriented development approach with rapid test cycles and direct stakeholder involvement (i.e., students, parents, teachers).