There isn’t voting in the model per se. It is a logical framework, which supposes that people have genuine subjective rankings of all possible outcomes. It is a separate matter to figure out how, in reality, outsiders would extract that information.
So I guess to be consistent with the “rational expectations” revolution (which we cover much later in the course), you would say that the people in Arrow’s model realize they are part of a society and that their individual preferences are being used as inputs for a broader judgment. But they are not allowed to lie about their preferences, if that’s what you’re getting at.