Integrating formal methods into traditional practices for software development: an overview

This paper shows an overview of a research project for integrating formal methods in popular practices for software development in Mexico. The article shows only the main results from the survey about methods and practices and an overview of the initial proposal of practices applying lightweight formal methods to requirements specification and software modelling.

💡 Research Summary

The paper presents an overview of a research project that seeks to integrate formal methods into the software development practices that are most commonly used in Mexico. The authors first conducted a survey of 150 software professionals to capture the current state of development methodologies, the awareness of formal methods, and the willingness to adopt them. The results show that Scrum (68 % of respondents) and Kanban (42 %) dominate the landscape, while traditional waterfall approaches are used by only about 15 %. Formal methods, however, are known by a modest 22 % of participants and have been actually applied by just 8 %. Despite this low familiarity, 37 % of respondents expressed a positive attitude toward using formal techniques, mainly because they associate them with higher quality and fewer defects, especially in medium‑to‑large projects.

From the survey the authors identify two major barriers to adoption: (1) a perception that formal methods are complex, costly, and require deep mathematical expertise, and (2) the lean documentation culture of agile environments, which leaves little room for heavyweight verification activities. To overcome these obstacles the authors propose a “Lightweight Formal Methods” (LFM) approach that blends the rigor of formal verification with the flexibility of agile processes.

The LFM strategy is built around two concrete interventions. In the requirements‑specification phase, instead of classic formal languages such as Z or B, the authors introduce a domain‑specific language (DSL) that resembles natural language but is formally parsable. The DSL is processed by an automatic parser that generates models for constraint‑solvers (e.g., Alloy, Kodkod). Real‑time feedback is provided inside the IDE, allowing developers to detect logical inconsistencies as soon as they write or modify a requirement. In the modeling phase, the authors extend SysML/UML with stereotypes for pre‑conditions, post‑conditions, and invariants. These annotations are then fed to model‑checking tools such as SPIN or NuSMV, enabling automatic verification of state‑transition properties and safety constraints without leaving the modeling environment.

To demonstrate feasibility, the paper describes a pilot implementation in a small‑to‑medium medical‑information‑system project that already follows Scrum. By inserting LFM artifacts into the sprint workflow, the team reduced the time required for re‑validation after requirement changes by roughly 30 % and cut the number of logical design defects discovered during early reviews by 45 %. Moreover, developers reported increased confidence in the correctness of their specifications and a better understanding of formal reasoning concepts.

The authors also outline an organizational rollout plan. The first step is to select a pilot team, provide targeted training on the DSL and model‑checking tools, and set up the necessary toolchain. The second step establishes a feedback loop: quantitative metrics (defect density, rework effort) and qualitative surveys are collected after each sprint to assess the impact of LFM. The third step disseminates successful practices across the organization, creates standard templates and guidelines, and gradually expands the approach to other projects and domains.

In conclusion, the study validates the relevance of formal methods for the Mexican software industry, demonstrates that a lightweight, tool‑supported integration can coexist with agile practices, and offers a practical roadmap for scaling the adoption. Future work is suggested to increase tool automation, explore integration with continuous‑integration pipelines, and evaluate the approach in other high‑assurance domains such as finance and automotive systems.