Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Aurora: Neuro-Symbolic AI Driven Advising Agent

Lorena Amanda Quincoso Lugones, Christopher Kverne, Nityam Sharadkumar Bhimani, Ana Carolina Oliveira, Agoritsa Polyzou, Christine Lisetti, Janki Bhimani

TL;DR

Aurora tackles the scalability and reliability gap in academic advising by integrating retrieval-augmented generation with strict symbolic reasoning over a BCNF-normalized catalog. Its modular design couples SQL-based pruning and Prolog enforcement with an instruction-tuned LLM to produce natural-language, policy-compliant degree plans with auditable provenance. The approach yields substantial gains in semantic alignment and precision/recall while delivering sub-second responses on commodity hardware, addressing both accuracy and practicality. This work demonstrates that grounding LLM outputs in verifiable curricular logic can transform AI-driven advising from a prototype to a trustworthy, scalable tool for higher education.

Abstract

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.

Aurora: Neuro-Symbolic AI Driven Advising Agent

TL;DR

Aurora tackles the scalability and reliability gap in academic advising by integrating retrieval-augmented generation with strict symbolic reasoning over a BCNF-normalized catalog. Its modular design couples SQL-based pruning and Prolog enforcement with an instruction-tuned LLM to produce natural-language, policy-compliant degree plans with auditable provenance. The approach yields substantial gains in semantic alignment and precision/recall while delivering sub-second responses on commodity hardware, addressing both accuracy and practicality. This work demonstrates that grounding LLM outputs in verifiable curricular logic can transform AI-driven advising from a prototype to a trustworthy, scalable tool for higher education.

Abstract

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.
Paper Structure (31 sections, 3 figures, 1 table, 1 algorithm)

This paper contains 31 sections, 3 figures, 1 table, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Aurora's Architecture
  4. PostgreSQL ER Knowledge Base Design
  5. Neuro-Symbolic Backend
  6. Intent, Filtering, and Symbolic Validation
  7. Chain-of-Thought Controller and Prompting
  8. From RAG to structured prompting.
  9. LLM Core Module
  10. Neuro-Symbolic Paradigm Classification
  11. Limitations of LLMs and Aurora’s Mitigations
  12. Experimental Setup
  13. Design Rationale
  14. Data and Simulation Environment
  15. Benchmark and Ground Truth
  16. ...and 16 more sections

Figures (3)

  • Figure 1: Aurora system overview. Green = knowledge base; blue = neuro-symbolic reasoning; red = Chain-of-Thought prompt segment. Solid arrows denote data flow; dashed line box show control or context boundaries.
  • Figure 2: Conceptual ER diagram. Purple = entities; green = relationships. Connectivity encoded in three types of links: 0..* ( e.g., a course can cover zero or many skills); 1..* (e.g., a program‐offering may include at least one course); 1..1 (e.g., every program‐offering is tied to exactly one major).
  • Figure 3: Per-query cosine similarity averaged over five runs.