Neuro-Symbolic AI: Foundations and Frontiers
DOI:
https://doi.org/10.46610/JoIDTA.2025.v02i03.001Keywords:
Cognitive intelligence, Deep learning, Explainability and trustworthiness, Hybrid systems, Knowledge representation, Learning and inference, Logic and reasoning, Machine learning, Meta-cognition, Neuro-symbolic AI, PRISMA, Systematic reviewAbstract
The field of Artificial Intelligence (AI) is currently in its "third AI summer," marked by significant advancements and the emergence of Neuro-Symbolic AI (NSAI). Traditional AI paradigms, such as Symbolic AI and Sub-Symbolic (deep learning), face distinct limitations: Symbolic AI struggles with real-world noise, while deep learning often lacks reasoning, interpretability, and robustness, leading to concerns about computational sustainability and limited human-AI collaboration. NSAI aims to bridge these divides by integrating logical reasoning with neural networks, fostering systems with enhanced explainability, trustworthiness, and data efficiency, ultimately striving for human-like cognitive capabilities.
This systematic review synthesizes NSAI progress from 2020-2024, highlighting key developments, methodologies, and applications, with a particular focus on healthcare. Findings indicate research concentration in learning and inference, logic and reasoning, and knowledge representation. However, critical gaps persist in explainability, trustworthiness, and especially Meta-Cognition, which is essential for self-monitoring and adaptation. Addressing these challenges, including knowledge representation complexity and the absence of standardized benchmarks, will be crucial. Advancing NSAI promises more autonomous, adaptable, reliable, and context-aware AI systems, with profound implications for areas like drug discovery and patient care.
References
S. Badreddine, A. d’Avila Garcez, L. Serafini, and M. Spranger, “Logic Tensor Networks,” Artificial Intelligence, vol. 303, p. 103649, Feb. 2022, doi: https://doi.org/10.1016/j.artint.2021.103649.
M. Drancé, M. Boudin, F. Mougin, and G. Diallo, “Neuro-symbolic XAI for Computational Drug Repurposing,” Proceedings of the 13th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management, 2021, doi: https://doi.org/10.5220/0010714100003064.
Kexin Yi and Jiajun Wu, “Neural-symbolic VQA: disentangling reasoning from vision and language understanding,” NIPS’18: Proceedings of the 32nd International Conference on Neural Information Processing Systems, 2018. https://dl.acm.org/doi/abs/10.5555/3326943.3327039
S.-I. Jang, Michael, and A. H. Thiery, “Explainable and Interpretable Diabetic Retinopathy Classification Based on Neural-Symbolic Learning,” arXiv (Cornell University), Jan. 2022, doi: https://doi.org/10.48550/arxiv.2204.00624.
A. Karim, M. Lee, T. Balle, and A. Sattar, “CardioTox net: a robust predictor for hERG channel blockade based on deep learning meta-feature ensembles,” Journal of Cheminformatics, vol. 13, no. 1, Aug. 2021, doi: https://doi.org/10.1186/s13321-021-00541-z.
S. Kim, J. Chen, T. Cheng, “PubChem 2023 update,” Nucleic Acids Research, vol. 51, no. D1, Oct. 2022, doi: https://doi.org/10.1093/nar/gkac956.
B. C. Colelough and W. Regli, “Neuro-Symbolic AI in 2024: A Systematic Review,” Arxiv.org, 2024. https://arxiv.org/abs/2501.05435
Garcez and L. C. Lamb, “Neurosymbolic AI: the 3rd wave,” Artificial Intelligence Review, Mar. 2023, doi: https://doi.org/10.1007/s10462-023-10448-w.
E. Ilkou and M. Koutraki, “Symbolic Vs Sub-symbolic AI Methods: Friends or Enemies?” 2020. Available: https://ceur-ws.org/Vol-2699/paper06.pdf
R. Maclin and J. W. Shavlik, “Refining Algorithms with Knowledge-Based Neural Networks: Improving the Chou-Fasman Algorithm for Protein Folding*,” Machine Learning Research Group Working Paper, 1992. https://ftp.cs.wisc.edu/machine-learning/shavlik-group/maclin.mlrgwp91.pdf
Michiel O. Noordewier and Geoffrey G. Towell, “Training knowledge-based neural networks to recognize genes in DNA sequences,” NIPS’90: Proceedings of the 4th International Conference on Neural Information Processing Systems, 1990. https://dl.acm.org/doi/abs/10.5555/2986766.2986838
R. Maclin and J. W. Shavlik, “Refining Domain Theories Expressed as Finite-State Automata,” Elsevier eBooks, pp. 524–528, Jan. 1991, doi: https://doi.org/10.1016/b978-1-55860-200-7.50107-0.
M. Sordo, H. Burton, and D. Watson, “KBANNs and the classification of,” 9th International Conference on Artificial Neural Networks: ICANN ’99, vol. 1999, pp. 982–987, Jan. 1999, doi: https://doi.org/10.1049/cp:19991240.
D. Hossain and J. Y. Chen, “A Study on Neuro-Symbolic Artificial Intelligence: Healthcare Perspectives,” Arxiv.org, 2025. https://arxiv.org/abs/2503.18213
J. Mao, C. Gan, P. Kohli, J. B. Tenenbaum, and J. Wu, “The Neuro-Symbolic Concept Learner: Interpreting Scenes, Words, and Sentences from Natural Supervision,” Arxiv:1904.12584, Apr. 2019, Available: https://arxiv.org/abs/1904.12584
J. Shi, H. Zhang, and J. Li, “Explainable and Explicit Visual Reasoning Over Scene Graphs,” Arxiv (Cornell University), Jun. 2019, doi: https://doi.org/10.1109/cvpr.2019.00857.
L. L. Custode, F. Mento, S. Afrakhteh, “Neuro-symbolic interpretable AI for automatic COVID-19 patient-stratification based on standardised lung ultrasound data,” Proceedings of Meetings on Acoustics, vol. 46, p. 020002, 2022, doi: https://doi.org/10.1121/2.0001600.
