Search and Filter

Aurora: Neuro-Symbolic AI Driven Advising Agent

Date

AI Overview

Why it matters

High student-to-advisor ratios strain higher education advising, limiting timely guidance and increasing graduation delays. Unconstrained language models often hallucinate courses or violate prerequisite rules, threatening policy compliance and student progress.

Study at a glance
  • Question: Can combining symbolic rule validation with language models produce accurate, policy-compliant, and scalable academic course recommendations?
  • Setting and data: Simulated undergraduate advising environment across four programs (CS, Data Science, IT, Product Management) using 20 benchmark queries.
  • Method: System development and benchmark evaluation comparing a hybrid neuro-symbolic pipeline against an unconstrained language model baseline.
  • AI tools: Aurora, DeepSeek-R1-Distill-Qwen-7B
Key findings
  • Aurora improved semantic alignment with expert advising answers from 0.68 to 0.93 compared to the raw language model baseline.
  • Aurora achieved course-level precision, recall, and F1 scores averaging 0.81 to 0.83, reaching perfect accuracy in about half of in-scope cases.
  • Aurora sustained sub-second average response times (0.71 seconds per query) on commodity hardware while reliably providing correct fallbacks for unanswerable prompts.

Summary drafted with AI from the full paper and may contain errors. See the full paper for details.

Academic advising in higher education is under severe strain, with advisor-to-student ratios commonly exceeding 300:1. These structural bottlenecks limit timely access to guidance, increase the risk of delayed graduation, and contribute to inequities in student support. We introduce Aurora, a modular neuro-symbolic advising agent that unifies retrieval-augmented generation (RAG), symbolic reasoning, and normalized curricular databases to deliver policy-compliant, verifiable recommendations at scale. Aurora integrates three components: (i) a Boyce-Codd Normal Form (BCNF) catalog schema for consistent program rules, (ii) a Prolog engine for prerequisite and credit enforcement, and (iii) an instruction-tuned large language model for natural-language explanations of its recommendations. To assess performance, we design a structured evaluation suite spanning common and edge-case advising scenarios, including short-term scheduling, long-term roadmapping, skill-aligned pathways, and out-of-scope requests. Across this diverse set, Aurora improves semantic alignment with expert-crafted answers from 0.68 (Raw LLM baseline) to 0.93 (+36%), achieves perfect precision and recall in nearly half of in-scope cases, and consistently produces correct fallbacks for unanswerable prompts. On commodity hardware, Aurora delivers sub-second mean latency (0.71s across 20 queries), approximately 83X faster than a Raw LLM baseline (59.2s). By combining symbolic rigor with neural fluency, Aurora advances a paradigm for accurate, explainable, and scalable AI-driven advising.