Evaluating Technical Manual Designs in AI Development Tasks Using QN-MHP-U

Traditional technical manual design largely relies on subjective experience and lacks quantitative methods for evaluating typographic and visual presentation strategies. As artificial intelligence (AI) technologies proliferate, the operational complexity of advanced systems escalates. In typical AI development tasks, the quality of technical manuals directly dictates the cognitive barriers encountered by novices. To address this issue, this study proposes a computational modeling approach based on the Queuing Network-Model Human Processor (Unified and Ease of Use, QN-MHP-U). The proposed method extends the computational logic of the baseline visual processing server (Server 1) to quantitatively model the effects of font size, line spacing, and text bolding on visual perception. In addition, a reading comfort metric C _read is introduced to evaluate subjective reading experience. Five technical manual layouts with different typographic configurations and text–image strategies were evaluated, among which three layouts (D1, D3, and D4) were further validated through empirical usability experiments. The extended model achieved high predictive accuracy for both task completion time ( R ²  = .93, RMSE = 42.67 s) and subjective reading comfort ( R ²  = .86, RMSE = 6.12). The results indicate that hierarchically structured procedural text organization, coordinated text–image layouts, and optimized typographic settings can substantially improve reading comfort and enhance task execution efficiency. Overall, this study provides an intelligent evaluation framework for technical manual design, enabling designers to optimize visual presentation strategies and reduce the cognitive load experienced by novice users.