There is a clear dispreference for using contractions in term papers and journal publications, and a clear preference for using 'em in ordinary conversation. There's an extremely strong dispreference for contracting future will when the point is to deny a claim that some event won't happen, so it would be bizarre to say "No, he'LL bring it back". In mixed style writing (e.g. SE), you'll find both outcomes, and I doubt that you will find a consistent pattern.
This paper looks at effect of word frequency on contraction, and happens to include ample references to quantitative studies and factors influencing contraction (phonology of preceding word, syntactic nature of following comp, word length of subject NP, information load). This blog might be informative as well, from a historical perspective (did we used to not contract in American English?), and there's a dissertation here on yet another aspect of contraction. I don't know of an experimental study that asks listeners to attitudinally judge spoken English samples as "educated", "stilted", "ignorant", "awkward", "pretentious", "sloppy" and so on, as a function of contracting vs. not contracting.