TRACE-CS: A Hybrid Logic-LLM System for Explainable Course Scheduling

Hybrid systems can improve AI trustworthiness: TRACE-CS demonstrates that combining traditional logic-based reasoning with LLMs produces "provably correct explanations" rather than relying on LLMs alone, which are prone to hallucinations. This approach is particularly valuable for high-stakes decisions like course scheduling where accuracy matters.

Explainability requires multiple layers: The system uses symbolic logic to ensure correctness and LLMs to translate those explanations into "user-friendly responses." This two-step process acknowledges that technical correctness alone isn't enough—AI systems must communicate their reasoning in ways people can actually understand and verify.

Symbolic reasoning provides accountability: By encoding scheduling constraints through logic-based techniques, TRACE-CS creates a verifiable foundation for decisions. This means stakeholders can trace why a particular schedule was chosen, rather than accepting an opaque algorithmic decision, which is essential for responsible deployment in educational institutions.

Natural language interfaces need guardrails: The paper shows that LLMs can help make complex systems more accessible, but only when constrained by logical frameworks. Without this structure, natural language interfaces risk introducing errors or biases into otherwise sound decision-making processes.