Towards a Perspective-Based Multi-Agent Architecture for Requirements Engineering of ML-Enabled Systems

• Jose M. C. Boaro PUC-Rio https://orcid.org/0000-0002-4456-9050
• Matheus P. Soranço de Carvalho PUC-Rio https://orcid.org/0009-0003-1892-5976
• Marcos Kalinowski PUC-Rio https://orcid.org/0000-0003-1445-3425

DOI: https://doi.org/10.5753/se4as.2026.32379

Resumo

[Context] Requirements Engineering (RE) for Machine Learning (ML)-enabled systems requires reconciling multiple, interdependent concerns, spanning system objectives, user experience, infrastructure, model, and data, that are difficult to capture through a single, monolithic specification process. Large Language Models (LLMs) show promising results in supporting RE tasks; however, when applied as general-purpose models to specify ML-enabled systems, they tend to produce shallow, monolithic specifications that ignore the dependencies among these concerns, while raising cost, privacy, and sustainability issues. [Goal] In this vision paper, we aim to contribute to AI-assisted requirements engineering of ML-enabled systems by discussing the feasibility of using Small Language Models (SLMs) to support this task. [Method] We ground our vision in PerSpecML, a perspective-based ML specification approach validated in academia and industry, whose five perspectives define the scope and boundaries of specialized agents. [Results] Our vision comprises a multi-agent architecture in which each agent is instantiated with an SLM fine-tuned to its corresponding perspective and coordinated by an orchestrator agent, under sequential, parallel, or hybrid topologies, with chain-of-thought reasoning intended to support reviewable rationale and human-in-the-loop review. [Conclusion] We acknowledge that this proposal still demands empirical evaluation, and we present it as a starting point for the community to discuss the feasibility of perspective-based multi-agent RE.

Palavras-chave: Requirements Engineering, Machine Learning-Enabled Systems, Multi-Agent Systems, Small Language Models, PerSpecML

Referências

Santos Alves, A. P., Kalinowski, M., Giray, G., Mendez, D., Lavesson, N., Azevedo, K., Villamizar, H., Escovedo, T., Lopes, H., and Gorschek, T. 2023. Status Quo and Problems of Requirements Engineering for Machine Learning: Results from an International Survey. In PROFES 2023. Springer, 159–177.

Belcak, P., Heinrich, G., Diao, S., Fu, Y., Dong, X., Muralidharan, S., Lin, Y. C., and Molchanov, P. 2025. Small Language Models are the Future of Agentic AI. arXiv preprint arXiv:2506.02153 (2025).

Guo, T., Chen, X., Wang, Y., Chang, R., Pei, S., Chawla, N. V., Wiest, O., and Zhang, X. 2024. Large Language Model Based Multi-Agents: A Survey of Progress and Challenges. In IJCAI 2024. 8048–8057.

Arora, C., Grundy, J., and Abdelrazek, M. 2024. Advancing Requirements Engineering Through Generative AI: Assessing the Role of LLMs. In Generative AI for Effective Software Development, Springer, 129–148.

Habibullah, K. M., Gay, G., and Horkoff, J. 2023. Non-Functional Requirements for Machine Learning: Understanding Current Use and Challenges Among Practitioners. Requirements Engineering 28, 2 (2023), 283–316.

He, J., Treude, C., and Lo, D. 2025. LLM-Based Multi-Agent Systems for Software Engineering: Literature Review, Vision, and the Road Ahead. ACM TOSEM 34, 5 (2025), 1–30.

International Organization for Standardization. 2023. ISO/IEC DIS 25059 – Quality Model for AI Systems. ISO.

Kalinowski, M., Mendez, D., Giray, G., Santos Alves, A. P., et al. 2025. Naming the Pain in Machine Learning-Enabled Systems Engineering. Information and Software Technology (2025).

Schwartz, R., Dodge, J., Smith, N. A., and Etzioni, O. 2020. Green AI. Communications of the ACM 63, 12 (2020), 54–63.

Villamizar, H., Kalinowski, M., Lopes, H., and Mendez, D. 2022. Towards Perspective-Based Specification of Machine Learning-Enabled Systems. In SEAA 2022. IEEE, 112–115.

Villamizar, H., Kalinowski, M., Lopes, H., and Mendez, D. 2024. Identifying Concerns When Specifying Machine Learning-Enabled Systems: A Perspective-Based Approach. Journal of Systems and Software 213 (2024), 112053.

Vogelsang, A. and Borg, M. 2019. Requirements Engineering for Machine Learning: Perspectives from Data Scientists. In REW 2019. IEEE, 245–251.

Wei, J., Wang, X., Schuurmans, D., Bosma, M., Ichter, B., Xia, F., Chi, E., Le, Q., and Zhou, D. 2022. Chain-of-Thought Prompting Elicits Reasoning in Large Language Models. In NeurIPS 2022, Vol. 35. 24824–24837.

Görer, B. and Aydemir, F. B. 2023. Generating Requirements Elicitation Interview Scripts with Large Language Models. In REW 2023. 44–51.

Arulmohan, S., Meurs, M.-J., and Mosser, S. 2023. Extracting Domain Models from Textual Requirements in the Era of Large Language Models. In MODELS-C 2023. 580–587.

Lutze, R. and Waldhör, K. 2024. Generating Specifications from Requirements Documents for Smart Devices Using Large Language Models (LLMs). In HCII 2024. Springer, 94–108.

Xu, X., Zhang, D., Liu, Q., Lu, Q., and Zhu, L. 2025. Agentic RAG with Human-in-the-Retrieval. In ICSA-C 2025. 498–502.