AITEE: Agentic Socratic Tutoring for Electrical Engineering

Fri, 20 Oct 2023 11:13:50 +0100

The AITEE project (Agentic Tutor for Electrical Engineering) develops an intelligent tutoring system for electrical engineering education. Students can ask questions about electrical circuits, which they may draw by hand or provide in digital form. The circuit serves as input for the tutor, which then discusses it with the student.

The goal is to support students in developing a deeper understanding of circuit analysis. Rather than presenting direct solutions, the tutor follows a Socratic approach: it offers hints, explanations and guiding questions that accompany students through their own learning process. This makes the tutor an additional resource that students can use independently and flexibly, and it encourages self-directed learning.

The system addresses three main challenges:

Recognition of hand-drawn circuits, including the validation of connections between components.
Retrieval of relevant lecture material. Since text-based similarity often fails for circuits, where the topology determines the solution method, the system uses graph neural networks to generate topology-based embeddings.
Response generation in a Socratic dialogue, while numerical results are verified by SPICE simulation instead of being computed by the language model.

Publications

C. Knievel, A. Bernhardt, C. Bernhardt: , IEEE Access, vol. 14, pp. 50109–50126, 2026. DOI: ·

Large Language Models | Christopher Knievel

AITEE: Agentic Socratic Tutoring for Electrical Engineering

Publications