Skip to main content Skip to navigation

Logical Modelling in CS Education: Bridging the Natural Language Gap

Project Overview

The document examines the role of generative AI, specifically through natural language processing (NLP), in enhancing computer science education by addressing the challenges of formal logic understanding. It presents a framework designed to facilitate vocabulary design tasks, enabling students to create appropriate vocabularies for formalizing real-world scenarios. The Iltis educational system serves as a practical example of this approach, allowing students to engage actively with complex topics by applying their vocabulary creations in logical modeling. The implementation of this framework aims to improve student engagement and comprehension in formal logic, a subject typically perceived as difficult, thereby fostering a deeper understanding of the connection between natural language and formal logical representations. Overall, the findings suggest that integrating generative AI tools in educational contexts can significantly enhance learning experiences and outcomes in computer science.

Key Applications

Iltis educational system for vocabulary design tasks

Context: Introductory logic courses for computer science students at Ruhr University Bochum

Implementation: Students use an online platform to design vocabularies for propositional and first-order logic, with NLP models providing feedback on their choices.

Outcomes: Increased student engagement and understanding of logical modeling. The framework has been evaluated with a substantial dataset, showing high classification accuracy for student attempts.

Challenges: Initial lack of authentic student data for model training and the complexity of accurately checking student vocabulary designs.

Implementation Barriers

Technological barrier

The need for accurate and immediate feedback mechanisms in educational support systems, particularly for vocabulary design tasks. Additionally, concerns arise regarding the resource intensity of larger models.

Proposed Solutions: Developing small fine-tuned models tailored for the specific educational context, which can be deployed locally by instructors to ensure both effective feedback and data sovereignty.

Project Team

Tristan Kneisel

Researcher

Fabian Vehlken

Researcher

Thomas Zeume

Researcher

Contact Information

For information about the paper, please contact the authors.

Authors: Tristan Kneisel, Fabian Vehlken, Thomas Zeume

Source Publication: View Original PaperLink opens in a new window

Project Contact: Dr. Jianhua Yang

LLM Model Version: gpt-4o-mini-2024-07-18

Analysis Provider: Openai

Let us know you agree to cookies