Abstract

Systems capability is the cognitive ability to plan and execute actions using systems-oriented reasoning. While essential for the design and development of complex systems, it manifests in large organizations primarily through communication and technical documentation, making it difficult to observe and measure at scale. This paper presents an automated method for inferring systems capability from natural language artifacts. We construct a feature-based representation that captures linguistic indicators of systems-oriented reasoning and use these features to train an ordinal logistic regression model on 75 graded systems engineering essay-type case reports, with expert-assigned quality grades serving as proxies for capability levels. The model is evaluated using a 60–20–20 train–validation–test split, achieving 53.3% test accuracy and a mean absolute error of 0.73 grade levels. Notably, 93.3% of predictions fall within ±1 grade level of the true assessments, indicating that specific linguistic patterns and systems attributes are strongly associated with higher systems capability. The proposed approach enables scalable assessment of systems capability with applications in engineering education, workforce development, and enterprise planning. Future work will refine the model and evaluate its performance in industrial settings.

IPC Classification

Keywords