PriseBot - A chatbot to assist in the development of iStar Extensions

In the world of software design, there is a long-standing challenge: how to explain not just what a computer program should do, but why it should do it. Engineers often use a method called goal-oriented requirements engineering, which treats software development as a series of objectives that people and systems must achieve together. This approach relies on a specific language called iStar, which uses diagrams to map out how different actors, like users or organizations, depend on one another to reach their goals. While this language is powerful, it has a limitation. As the needs of modern technology grow more complex, the standard rules of iStar sometimes fall short. To fix this, experts create "extensions," which are custom additions to the language that allow it to describe new, specific situations. However, creating these extensions is a difficult task that requires deep expertise and a strict, step-by-step process to ensure the new rules fit correctly with the old ones.

This is where a new tool called PriseBot enters the story. Researchers from the Federal University of Ceará in Brazil developed this chatbot to act as a guide for anyone trying to create these custom extensions. Instead of forcing users to read through dense manuals or complex flowcharts, PriseBot allows them to ask questions in plain language, just as they would speak to a colleague. The system is built on a foundation of existing knowledge about the iStar language and the specific process for making extensions, known as PRISE. By combining a conversational interface with a mechanism that retrieves precise information from a database of scientific articles and technical documents, the bot helps users navigate the complicated steps of designing new software rules without getting lost.

The researchers did not just build the tool; they put it to the test in two distinct ways to see if it actually worked. First, they watched a student with limited experience use PriseBot to create a brand-new extension for modeling chatbots. The student relied on the bot throughout the entire process, from the initial idea to the final design. The student reported that the tool was incredibly useful, noting that it made the complex process feel manageable and provided clear, structured answers that saved time. The bot successfully guided the student through every stage, helping them define new concepts and check their work against the required standards.

To see if this success was a fluke or something broader, the team then surveyed twenty-three experienced researchers who specialize in creating iStar extensions. These experts were asked to evaluate the bot's potential usefulness, ease of use, and ability to improve their productivity. The results were overwhelmingly positive. The vast majority of these seasoned professionals agreed that the bot would be helpful in their work, that it would make their tasks faster, and that it would increase their chances of successfully finishing a project. They found the interface intuitive and the learning curve gentle. When compared to general-purpose artificial intelligence tools that are not trained on this specific technical knowledge, PriseBot performed significantly better, answering every question correctly while the other tools failed to provide accurate answers on the same topics.

The study also demonstrated the bot's practical value by using it to create a real-world extension called iStar4Bot. This new addition to the iStar language was designed specifically to help engineers model the requirements for chatbots, a rapidly growing field. The extension introduced new concepts, such as "intent" and "context," which allow for a more precise description of how a chatbot interacts with a human. The fact that a working, validated extension could be produced with the help of the chatbot suggests that the tool is not just a theoretical experiment but a functional aid that can bridge the gap between complex theory and practical application.

Ultimately, the research shows that bringing conversational technology into the specialized field of software engineering can demystify complex processes. By wrapping a rigorous, step-by-step methodology in a friendly, interactive interface, the researchers have created a way for both beginners and experts to engage with advanced software design concepts more effectively. The findings suggest that when artificial intelligence is carefully trained on specific, high-quality knowledge and guided by a clear process, it can become a reliable partner in solving difficult engineering problems, making the creation of better software systems more accessible to a wider range of people.