The increasing complexity of organizational, environmental, and operational decision environments requires artificial intelligence systems that can integrate heterogeneous information, recognize previously unseen situations, and support decisions under uncertainty. Conventional AI pipelines frequently depend on predefined classes, static datasets, and narrowly specified prediction objectives, limiting their ability to adapt when decision contexts evolve. This paper proposes a conceptual Semantic AI Architecture for Sustainable and Intelligent Decision-Making that combines semantic representation, open-world learning, incremental learning, attention-based representation, self-supervised feature extraction, and information-theoretic reasoning. The architecture is developed through a structured synthesis of the supplied literature and is positioned as a research-oriented framework rather than an empirical benchmark. The analysis demonstrates that sustainability-oriented decision intelligence requires more than predictive accuracy: it requires contextual interpretation, novelty awareness, continuous learning, and mechanisms for integrating multimodal evidence. Open-world recognition provides a foundation for identifying unknown situations, while incremental learning enables adaptation without complete retraining. Transformer-based representations and self-supervised learning can improve the semantic quality of continuously acquired data, while information theory provides a theoretical basis for evaluating information relevance and uncertainty. The proposed architecture therefore connects semantic interpretation with adaptive AI infrastructure and sustainable decision processes. The study identifies architectural advantages, operational trade-offs, and limitations, particularly concerning semantic consistency, computational cost, uncertainty management, and evaluation of genuinely unknown decision states.
Semantic AI Architecture for Sustainable and Intelligent Decision-Making
DOI:
Abstract
References
Beddiar, D. R., Nini, B., Sabokrou, M., & Hadid, A. (2020). Vision-based human activity recogni-tion: a survey.Multimed Tools Appl,79(41), 30509–30555.
Belouadah, E., Popescu, A., & Kanellos, I. (2021). A comprehensive study of class incrementallearning algorithms for visual tasks.Neural Networks,135, 38–54.
Bendale, A., & Boult, T. (2015). Towards Open World Recognition. InConference on ComputerVision and Pattern Recognition, pp. 1893–1902.
Bertasius, G., Wang, H., & Torresani, L. (2021). Is Space-Time Attention All You Need for VideoUnderstanding?. InProceedings of the 38th International Conference on Machine Learning,pp. 813–824. PMLR. ISSN: 2640-3498.
Boult, T. E., Cruz, S., Dhamija, A. R., Gunther, M., Henrydoss, J., & Scheirer, W. J. (2019). Learn-ing and the Unknown: Surveying Steps toward Open World Recognition.Proceedings of theAAAI Conference on Artificial Intelligence,33(01), 9801–9807. Number: 01.
Boult, T., Grabowicz, P., Prijatelj, D., Stern, R., Holder, L., Alspector, J., Jafarzadeh, M. M., Ahmad,T., Dhamija, A., Li, C., Cruz, S., Shrivastava, A., Vondrick, C., & Walter, S. (2021). Towardsa Unifying Framework for Formal Theories of Novelty.Proceedings of the AAAI Conferenceon Artificial Intelligence,35(17), 15047–15052. Number: 17.
Brighi, M., Franco, A., & Maio, D. (2021). ActivityExplorer: A semi-supervised approach to dis-cover unknown activity classes in HAR systems.Pattern Recognition Letters,151, 340–347.
Carreira, J., Noland, E., Banki-Horvath, A., Hillier, C., & Zisserman, A. (2018). A Short Note aboutKinetics-600.arXiv:1808.01340 [cs],””. arXiv: 1808.01340.
Carreira, J., Noland, E., Hillier, C., & Zisserman, A. (2019). A Short Note on the Kinetics-700Human Action Dataset.arXiv:1907.06987 [cs],””. arXiv: 1907.06987.
Chalmers, D. J. (2006). Strong and weak emergence. InThe Re-Emergence of Emergence: TheEmergentist Hypothesis from Science to Religion, pp. 244–415. OUP Oxford.
Cover, T. M., & Thomas, J. A. (1991).Elements of Information Theory. Wiley Series in Telecom-munications. John Wiley & Sons, Inc., New York, USA.
Dhamija, A. R., Ahmad, T., Schwan, J., Jafarzadeh, M., Li, C., & Boult, T. E. (2021). Self-Supervised Features Improve Open-World Learning.arXiv:2102.07848 [cs],””. arXiv:2102.07848.
Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T., Dehghani,M., Minderer, M., Heigold, G., Gelly, S., Uszkoreit, J., & Houlsby, N. (2022). An Image isWorth 16x16 Words: Transformers for Image Recognition at Scale. InInternational Confer-ence on Learning Representations.
K. K. Goyal, "Semantic AI Infrastructure for Sustainable Decision Intelligence," 2025 8th International Conference on Informatics and Computational Sciences (ICICoS), Semarang, Indonesia, 2025, pp. 434-439, doi: 10.1109/ICICoS68590.2025.11329920.