Revista ELECTRO
Vol. 47 – Año 2025
Artículo
TÍTULO
Arquitectura Neuro-Simbólica para Razonamiento Causal Explicable
AUTORES
Márquez-Gutiérrez, P.R.; García-Mata, C.L.; Acosta-Cano de los Ríos, P.R.; Baray-Arana; R.E.; Robledo-Vega, I.
RESUMEN
La convergencia entre métodos neuronales y simbólicos ha abierto nuevas oportunidades en inteligencia artific ial, particularmente para lograr modelos que ofrezcan tanto alto desempeño como explicabilidad. En este artículo revisamos el estado del arte en cómputo neuro-simbólico y proponemos un marco híbrido orientado a incorporar razonamiento causal explícito en e scenarios de toma de decisión. Tras exponer motivaciones y desafíos, describimos una arquitectura que combina un “backbone ” neuronal (Transformer o CNN según el dominio) con un módulo simbólico basado en lógica causal probabilística (p. ej., ProbLog). Pres entamos un método de entrenamiento conjunto que ajusta pesos neuronales y umbrales de discretización de predicados, permitiendo que el componente simbólico genere cadenas de inferencia legibles. Se valora la propuesta en dos dominios: diagnóstico médico (detección de causas de síntomas) y planificación en robótica de asistencia (tareas con reglas causales). Los resultados muestran mejoras tanto en métricas predictivas como en fidelidad causal y comprensibilidad de explicaciones, en comparación con modelos p uramente neuronales o simbólicos. Concluimos destacando el potencial de arquitecturas neuro-simbólicas causales y proponiendo líneas futuras de investigación.
Palabras Clave: IA neuro-simbólica, razonamiento causal, explicabilidad, aprendizaje profundo, ProbLog, BioBERT, ResNet, lógica probabilística, diagnóstico médico, robótica de asistencia.
ABSTRACT
The convergence between neural and symbolic methods has opened new opportunities in artificial intelligence, particularly for achieving models that offer bo th high performance and explainability. In this article, we review the state of the art in neuro-symbolic computing and propose a hybrid framework designed to incorporate explicit causal reasoning into decision-making scenarios. After presenting motivation s and challenges, we describe an architecture that combines a neural backbone (Transformer or CNN, depending on the domain) with a symbolic module based on probabilistic causal logic (e.g., ProbLog). We introduce a joint training method that adjusts neural weights and predicate discretization thresholds, enabling the symbolic component to produce readable inference chains. The proposal is evaluated in two domains: medical diagnosis (detection of symptom causes) and planning in assistive robotics (tasks with causal rules). Results show improvements in both predictive metrics and causal fidelity and in the comprehensibility of explanations, compared to purely neural or symbolic models. We conclude by highlighting the potential of causal neuro-symbolic architec tures and proposing future research directions.
Keywords: neuro-symbolic AI, causal reasoning, explainability, deep learning, ProbLog, BioBERT, ResNet, probabilistic logic, medical diagnosis, assistive robotics.
REFERENCIAS
[1] LeCun, Y., Bengio, Y., & Hinton, G., Deep learning. Nature, 521 (7553), 436 –444, 2015. https://doi.org/10.1038/nature14539
[2] Rudin, C., Stop explaining black box m achine learning models for high stakes decisions and use interpretable models instead. Nature Machine Intelligence, 1 (5), 206 –215, 2019. https://doi.org/10.1038/s42256-019-0048-x
[3] Besold, T. R., d’Avil a Garcez, A. S., Bader, S., Bowman, H., Domingos, P., Hitzler, P., Valiant, L., Neural-Symbolic Learning and Reasoning: A Survey and Interpretation. Journal of Artificial Intelligence Research, 61, 491 –592, 2017. https://doi.org/10.1613/jair.5219
[4] Garcez, A. d’Avila, Lamb, L. C., & Gabbay, D. M., Neural-Symbolic Cognitive Reasoner: Integrating Reasoning, Learning, and Knowledge Representation. Springer International Publishing, 2019.
[5] Pearl, J., & Mackenzie, D., The Book of Why: The New Science of Cause and Effect. Basic Books, 2018.
[6] Glymour, C., Zhang, K., & Spirtes, P. (2019). Review of Causal Discovery Methods Based on Graphical Models. Frontiers in Genetics, 10, 524. https://doi.org/10.3389/fgene.2019.00524
[7] He, K., Zhang, X., Ren, S., & Sun, J., Deep Residual Learning for Image Recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR’16), pp. 770-778, 2016. https://doi.org/10.1109/CVPR.2016.90
[8] Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., … Polosukhin, I., Attention Is All You Need. In Proceedings of the 31st International Conference on Neural Information Processing Systems (NeurIPS 2017) (pp. 6000 –6010), 2017.
[9] Hannun, A., Case, C., Casper, J., Catanzaro, B., Diamos, G., Elsen, E., … Ng, A. Y., Deep Speech: Scaling up end-to-end speech recognition. arXiv preprint arXiv:1412.5567, 2019.
[10] Lundb erg, S. M., & Lee, S.-I., A Unified Approach to Interpreting Model Predictions. In Advances in Neural Information Processing Systems 30 (NeurIPS 2017) (pp. 4765 –4774) , 2017.
[11] Ribeiro, M. T., Singh, S., & Guestrin, C., “Why Should I Trust You?” Explaining the Predictions of Any Classifier. In Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD ’16) (pp.1135 –1144), 2016. https://doi.org/10.1145/2939672.2939778
[12] Richardson, M., & Domingos, P., Markov Logic Networks. Machine L earning, 62 (1–2), 107 –136, 2006. https://doi.org/10.1007/s10994-006-5833-1
[13] Pearl, J., Causality: Models, Reasoning, and Inference (2nd ed.). Cambridge University Press, 2009.
[14] Serafini, L., & Garcez, A. d’Avila, Logic Tensor Networks: Deep Learning and Logical Reasoning from Data and K nowledge. In Proceedings of the 25th International Joint Conference on Artificial Intelligence (IJCAI’16) (pp. 3342 –3348), 2016.
[15] Rocktäschel, T., & Riedel, S., End-to-End Differentiable Proving. In Advances in Neural Information Processing Systems 30 (NeurIPS 2017) (pp. 3788 –3800), 2017.
[16] Evans, R., & Gre fenstette, E., Learning Explanatory Rules from Noisy Data. In Proceedings of the 2018 Conference on Neural Information Processing Systems (Ne urIPS 2018) (pp. 6608 –6618), 2018.
[17] Manhaeve, R., Dumancic, S., Kimmig, A., Demeester, T., & De Raedt, L., DeepProbLog: Neural Probabilistic Logic Programming. In Proceedings of the 32nd AAA I Conference on Artificial Intelligence (AAAI ’18) (pp. 3756 –3764), 2018.
[18] Mao, J., Gan, C., Kohli, P., Tenenbaum, J. B., & Wu, J., The Neuro-Symbolic Concept Learner: Interpreting Scenes, Words, and Sentences From Natural Supervision. In Proceedings of the International Conference on Learning Representations (ICLR 2019).
[19] Kocaoglu, M., Shabanian, S., & Van der Schaar, M., CausalGAN: Learning Causal Implicit Generative Models with Adversarial Training. arXiv preprint arXiv:1805.08795, 2018.
[20] Liang, Y., Causal VAE: Disentangled Representation Learning with Causal Structures. In Proceedings of the 39th International Conference on Machine Learning (ICML 2022).
[21] Yang, Q., Barrault, L., & de Beauregard, R., Causal Heteroencoder: Learning Disenta ngled Representations with Causal Structures. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2022) (pp. 12840 –12849), 2022.
[22] Zhang, K., Chen, H., Wang, Y., & Glymour, C., Towards Neuro-Symbolic Causal AI: Le arning to Infer Counterfactuals with Symbolic Rules. In Proceedings of the 37th AAAI Conference on Artificial Intelligence (AAAI 2023) (pp. 14321 –14328), 2023.
[23] Lee, J., Yoon, W., Kim, S., Kim, D., Kim, S., So, C. H., & K ang, J., BioBERT: a pre‐trained biomedical language representation model for biomedical text mining. Bioinformatics, 36 (4), 1234 –1240, 2020. https://doi.org/10.1093/bioinformatics/btz682
[24] De Raedt, L., Kimmig, A., & Toivonen, H., ProbLog: A Probabilistic Prolog and its Application in Link Discovery. In Proceedings of the 20th International J oint Conference on Artificial Intelligence (IJCAI’07) (pp. 2468 –2473). Morgan Kaufmann, 2007.
[25] Johnson, A. E. W., Pollard, T. J., Shen, L., Lehman, L.-W. H., Feng, M., Gha ssemi, M., … Mark, R. G., MIMIC-III, a freely accessible critical care database. Scientific Data, 3, 160035, 2016. https://doi.o rg/10.1038/sdata.2016.35
[26] Shimizu, S., Hoyer, P. O., Hyvärinen, A., & Kerminen, A., A Linear Non-Gaussian Acyclic Model for Causal Discovery. Journ al of Machine Learning Research, 7, 2003 –2030, 2006.
[27] Spirtes, P., Gly mour, C., & Schei nes, R., Causation, Prediction, and Search (2nd ed.). MIT Press, 2000.
CITAR COMO:
Márquez-Gutiérrez, P.R.; García-Mata, C.L.; Acosta-Cano de los Ríos, P.R.; Baray-Arana; R.E.; Robledo-Vega, I., "Arquitectura Neuro-Simbólica para Razonamiento Causal Explicable", Revista ELECTRO, Vol. 47, 2025, pp. 161-176.
VERSIÓN PDF