Well-being can reflect people’s psychological conditions and be used alongside physiological parameters to evaluate patients’ physical and mental health. The modern medical environment increasingly incorporates digital carriers, human–computer interaction devices, sensible spaces, and the execution of suitable algorithms. Slow design in healthy human–computer interaction is often used to reflect people’s dependence on or support from behaviors or objects, promoting the stability of behaviors as well as meaningful and positive changes. Therefore, in this study, we propose a slow sensing model, develop a Slow Well-Being Gardening system, and use it to evaluate behavioral data from radiation therapy patients during treatment sessions and horticultural therapy. This study is based on SENS and slow design, setting the hospital lounge as a sensible space and establishing a sensor system. After a 10-day inspection, the process was evaluated and verified. Ultimately, data from facial detection (smile) and HRV showed that the patients in the experimental group experienced a significant improvement in their well-being, feeling better than those in the control group who maintained the most common state in normal treatment. Therefore, it can be inferred that the Slow Well-Being Gardening model is indeed valid and can be further developed.

In contemporary complex engineering projects, the core challenge of cross-disciplinary collaboration extends beyond technical information exchange to encompass the coordination and evolution of heterogeneous semantic interpretations through interaction. This study systematically investigates the semantic coordination dynamics between engineering and design professionals during early-stage collaboration, proposing a computational framework for tracking semantic evolution. A system comprising three interconnected modules—semantic externalization, visual mediation, and computational assessment—was developed to operationalize an iterative reinterpretation workflow. Participants’ keyword-based descriptions were transformed into traceable semantic vectors, and two coordination metrics were introduced: the Semantic Convergence Index (SCI) and the Cross-disciplinary Alignment Index (CAI). An empirical study was conducted with four mixed-discipline teams, each consisting of three engineering and three design participants, addressing the same design task. AI-generated images served as visual mediators to trigger cross-disciplinary reinterpretation across iterative cycles. Results demonstrate that semantic convergence and trajectory stability constitute analytically distinct coordination dimensions: stable interaction dynamics do not necessarily lead to semantic convergence, and convergence may emerge through multiple temporal pathways, including monotonic, delayed, and non-monotonic patterns. Two coordination patterns were observed within the present dataset—stable convergence and stable non-convergence—suggesting, within this exploratory context, that shared visual representations may not automatically produce shared understanding. This study provides engineering informatics with a diagnostic computational tool for monitoring coordination states under conditions of semantic heterogeneity and information incompleteness, establishing theoretical and methodological foundations for developing semantically aware decision support systems for early-stage design.