Requirements Elicitation for Prototype-driven AI Engineering: a Case Study in Police Report Generation

Abstract

[Context and Motivation] Systems containing AI components introduce novel challenges that increase the complexity of systems and their corresponding development processes. [Question / Problem] Traditional requirements engineering practices are unable to produce input for the development of AI-based systems in a way that covers all necessary aspects to develop responsible, compliant and safe solutions. [Principal ideas / Results] We developed PRE4AIM, a requirements elicitation method that enables the systematic inclusion of the required multidisciplinary perspectives needed for the development of AI-based systems, and applied this in a technical action research at the Netherlands Police. We were able to identify 23 requirements and were able to further enrich them by distinguishing obstacles, opportunities or possible solutions to satisfy them. [Contribution] Our method provides a novel approach for the elicitation of requirements for AI-based systems, that allows for improved domain understanding, risk identification, project planning and created meaningful input for future development of the current prototype.