Abstract
Current aging-friendly interaction design risk assessment relies heavily on expert judgment and mean-based indicators, failing to capture elderly users' behavioral nonlinear fluctuations, long-tail extreme risks, and temporal mutations resulting in unclear prudent boundaries for key parameters (size, spacing, font, feedback rhythm). A risk assessment model integrating non-parametric estimation, wavelet analysis, and prudence evaluation is proposed. Kernel density estimation constructs interaction behavior risk probability distributions; wavelet transform identifies temporal risk evolution; Value-at-Risk (VaR) establishes prudent boundaries; Monte Carlo simulation minimizes risk in design parameter space. Empirical tests were conducted on 20 elderly participants using an induction cooker's intelligent touch interface, collecting interaction data across parameter combinations. The model effectively identifies high-risk design zones. Recommended parameters (button width ≥ 16 mm, font size ≥ 14pt, response delay ≤ 500 ms) significantly reduce misoperation rates and task time, while improving user satisfaction and safety. Compared to traditional mean-based methods, this model better captures behavioral trends and controls extremes, suiting high-uncertainty user groups. It provides a "behavior modeling risk identification parameter feedback" closed-loop for aging-friendly design, expanding data-driven, prudent risk control in design science with value for multimodal interactions and intelligent elderly care.