*1.1. Survey-Based Data Collection*

Face-to-face interviews and workshops are the most common practices to collect opinions of the target groups. When it is impossible or too expensive to interview respondents through the direct meeting, online surveys might be especially helpful [10,11]. However, survey-based preference elicitation should be carefully organized, since criteria weighting results might be meaningfully affected by the survey structure, construction of the questions, or even by the visual means of the measurement scales. When contrasting unrelated questions are assessed, different survey items should be presented on separate pages [12,13]. If the respondents are asked to give ratings concerning the different aspects of a single latent variable, all the survey items should be intentionally presented on the same (web) page [14]. This particular way to present multiple, related items is called semantic differentials [15].

Semantic differentials are typically expressed as the matrix questions, where preferences are presented on the matrix side, and the response scale is presented on the top of it. Due to the ability to place multiple estimates on a single page, these structures are commonly used to collect public opinions on the quality, satisfaction, and the importance of the analyzed items [14]. Moreover, since humans are much better at making comparative judgments than the absolute ones [16,17], matrix questions might be valuable to increase the accuracy of the direct weighting techniques.
