Previous studies on predicting drivers’ proneness to violations and accidents emphasize demographic factors as significant predictors but overlook the link between individual risk preferences and driving safety. Our research aims to develop a scale assessing individuals’ risk attitudes, assuming risk preference is crucial for driving safety, particularly in conditionally automated driving (CAD). As an initial step, following Jian et al.’s (1988) questionnaire development methodology, we conducted a sentence elicitation study. Sentences relevant to risk attitude assessment were extracted from pertinent regulations. Three human factors experts reviewed the sentences, added additional ones, and evaluated them. Some of the evaluated sentences were compared pairwise to construct a dissimilarity matrix. A clustering analysis identified key dimensions of risk preference, forming the foundation for developing risk attitude assessment scale. This scale will help identify high-risk drivers, enabling targeted interventions such as tailored training programs and CAD interface modifications.