Electronic decision support for vulnerable-pedestrian crossings requires heterogeneous factors, including crossing intention, individual vulnerability, traffic conditions, and roadside facilities, to be represented on a common scale. Fixed-weight multicriteria ranking cannot adapt its evaluation priorities to changes in scenario risk. This study proposes an electronic decision-support method that combines an intention-probability-driven risk model with risk-adaptive TOPSIS. The model integrates crossing-intention probability, vulnerability coefficient, traffic density, environmental adversity, and facility deficiency. A convex combination of baseline and scenario-dependent exponential weights produces a composite risk score, which continuously interpolates the TOPSIS criterion weights. The method ranks three intervention options and provides traceable intermediate results for human confirmation. In six reproducible simulated scenarios, the proposed method achieved a policy-level consistency rate of 100.0%, exceeding equal-weight and fixed-weight TOPSIS by 66.7 and 83.3 percentage points, respectively. Under ±10% weight perturbations, the mean preferred-option retention rate was 92.3%, and one risk evaluation and ranking required 31.8 ± 0.7 microseconds. The results show that the method adjusts evaluation priorities with scenario risk and generates intervention recommendations consistent with the predefined response level.