Evaluation of thermal satisfaction scales through the analysis of sensation and preference votes

In this work, the analysis of thermal satisfaction votes expressed by students during regular classes was conducted to point
out (i) the capability of predicting satisfaction votes starting from sensation and preference votes and (ii) the correspondence
between two evaluation scales which are commonly used in the assessment of satisfaction. An experimental campaign in 50
classrooms in central Italy was carried out between 2020 and 2022 through the administration of questionnaires while
monitoring indoor environmental data. Questions about thermal sensation, preference and satisfaction were addressed.
While a 7-point-scale was used for thermal sensation and preference (-3 to +3), satisfaction was investigated using a 4-point,
one-pole-scale (0 to 3) as well as a dichotomic scale (0 to 1), with 0 as “satisfied” in both cases. First, classification trees
were used to train a model having thermal sensation and preference votes as independent variables, and satisfaction
expressed in the dichotomic scale as the dependent variable. The model was rendered as a two-entry color-coded matrix to
map the probability of satisfaction votes starting from given sensation and preference votes. Then, such model was used to
predict dichotomic satisfaction for the samples in which satisfaction vote was cast on the 4-point-scale. The comparison
between the predicted (0 to 1) and the observed (0 to 3) satisfaction votes highlighted that the “1” votes expressed in the 4-
point-scale correspond in 62% of the cases with thermal satisfaction. This is somewhat not consistent with the definition of
the scale, that assumes the 0 pole as absence of attribute, and the remaining votes as gradient degrees of dissatisfaction.
Extensions of this work will focus on (i) including environmental conditions as confounding factors (ii) extending such
analysis to the IAQ, visual and acoustic domains.