AI Breakthrough at Airlangga University Promises Faster, More Creative Software Development
Table of Contents
A new artificial intelligence (AI) method developed by researchers at Airlangga University (UNAIR) is poised to revolutionize how software features are conceived, offering a faster and more structured choice to traditional brainstorming sessions. The system leverages “service facts” – data from a service registry – across different industries to generate innovative solutions that are both relevant and previously unimagined.
Rethinking Requirements Engineering with AI
The findings, published in the IAES International Journal of Artificial Intelligence in December 2025, present a novel approach to requirements engineering, a critical phase in software development frequently enough constrained by time and resources. This AI-driven method aims to enrich this process by providing a constant stream of fresh ideas.
The system operates in two distinct phases. The first phase focuses on building “creativity resources.” This involves grouping service descriptions using techniques like TF-IDF, cosine similarity, and K-medoid, followed by the automated extraction of “service capabilities” – defined as verb-noun pairings – through dependency parsing utilizing Universal and Stanford Dependencies.
The second phase than matches a developer’s initial requirements against a knowledge base constructed using semantic similarity (including the Wu-Palmer method) and three structured creativity techniques: exploration (drawing analogies across domains), change (relaxing constraints), and combination (integrating new elements). This process ultimately generates a range of alternative solutions for consideration.
Scale and Speed: A Cross-Domain Idea Bank
The advantage of this approach lies in its ability to scale and operate with remarkable speed.In a feasibility test, the research team processed data from over 10,000 services across 10 different domains sourced from ProgrammableWeb. They successfully extracted functional capabilities from 98.8% of the descriptions, creating a readily available, cross-domain idea bank. This allows the system to “offer” alternative features that achieve the same objective but through potentially more inventive methods.
AI vs. Brainstorming: A Matter of originality
The quality of the AI-generated ideas was rigorously tested and compared to those produced through traditional brainstorming.While brainstorming sessions yielded a slightly higher average total creativity score (14.86) compared to the AI (13.57) in an experiment with seven participants,the AI demonstrably excelled in originality. “AI can generate alternative ideas instantly without the need for several rounds of discussion,” a senior official stated.
A case study involving an “order processing” flow highlighted this strength.The AI suggested changing “Create order” to “Offer order,” recommending that the company solicit bids from multiple suppliers simultaneously based on a target price, rather than committing to a single vendor. This alternative was deemed more creative, relevant, and useful.
Beyond Software: Expanding the Potential of AI-Driven Creativity
The implications of this research extend far beyond software engineering. The method could be integrated into existing development tools as a standalone interface or plugin for requirements platforms and Integrated Development Environments (IDEs). Furthermore, the same principles could be applied to product design, marketing, and business model innovation – provided structured knowledge components are available for recombination.
Researchers acknowledge limitations related to the quality and completeness of existing service descriptions. Future work will focus on enhancing the diversity and quality of solutions and extending the request of the method to later stages of development, including prototyping and testing.
The research was conducted by Faisal Fahmi (UNAIR),Feng-Jian Wang (National Yang-Ming Chiao-Tung University),and Hema Subramaniam (University of Malaya).
