Evaluating Generative AI as a Human Factors Requirements-Based Review System of Spaceflight Ground Displays

Abstract

Control displays provide monitoring, control, and commanding capabilities during prelaunch and launch countdown operations. These displays are evaluated against a set of design standards during development. Generative artificial intelligence systems offer an alternative to human-only display evaluations, allowing for a quick assessment of requirements compliance against approved standards. The present research evaluates the capabilities of GenAI systems for assessing launch control displays against a set of real-world usability and design standards. A custom GenAI implementation was developed and provided with a set of FAA and NASA design requirements. Additionally, the custom implementation was provided with any necessary context information to understand the operation of the three evaluated displays. Results show that while the display evaluator varied in its capabilities depending on the specific standards it was assessing, overall accuracy was very good. The findings suggest that GenAI systems are viable first-pass evaluators in requirements-based human factors review workflows.