polls represent emotions.
It seems to me that polls probably capture more than emotions. A given individual's assessment of Trump's job performance may very often depend on that individual's political ideology, policy preferences, and even some objective understanding of their own life situation, i.e. you might expect approval numbers to go up if more people perceive themselves to be economically better off. There's an emotional component to all of that, but it's not all emotion unless you think the people who respond to the polls make these assessments purely emotionally, which seems unlikely.
I'm not sure polls ought to be that
important to anyone -- even more so any individual poll vs. some aggregate. But they are somewhat useful and occasionally interesting measures of public opinion. The funny thing, to me, is that they are mostly useful by way of providing context to historical assessments, or in other critical applications, where you're more interested in knowing just what happened rather than what
should happen.
