Requirements Conflicts Detection: Advancing with Conversational AI

Conversational AI, which includes technologies like chatbots and virtual assistants facilitate Crowd-Based Requirements Engineering (CrowdRE) by streamlining the process of gathering and analyzing requirements from a large and diverse group of stakeholders through conversations. These technologies offer significant advantages, but they also present unique challenges, especially when dealing with conflicting requirements of crowds with different goals and needs. Such conflicts can lead to inconsistencies in system design, causing project delays, increased costs, and potential failures. This study introduces a novel solution to manage requirement conflicts by leveraging a Conversational AI developed with the open-source Rasa framework. The proposed system is designed to detect conflicts in real-time during stakeholder conversations. The study conducted an experiment to evaluate Conversational AI performance compared with requirements engineers using four different datasets. The preliminary evaluation of the Conversational AI shows its efficacy in real-time conflict detection. The analysis implies no significant difference in mean performance between the Conversational AI and the requirements engineers in detecting requirements conflicts. The initial results are promising for a user-friendly and efficient method for instantaneous conflict detection, although further training and evaluation in operational settings are needed. This opens avenues for conflict detection of crowd-based requirements and conflict resolution using Conversational AI, to enhance the quality and success rates of software projects.