Research papers I like are entertaining, make me think, and convey lots of new information. I wonder how does one go about writing a paper with these properties.
To be entertaining the paper has to be informal, to tell me about the hopes, the dreams, and the nightmares of the author. I want to hear the author saying the joy he felt when reading some obscure but beautiful result. I don't want a paper that is dead and puts me to sleep. I like papers that use small tricks to make me think. Mathematics papers in particular are very good at this. They keep using this trick of presenting a theorem before the proof. As a result, I read the theorem, try to prove it myself, fail, read the proof, and realize how stupid I am. That really makes me feel good. Finally, it's quite saddening to read most pieces of literature since they are such a waste of time. Umberto Eco said that ``Three lines of Pascal say more than three hundred pages of a long and tedious treatise on morals and metaphysics.'' Of course, it's not clear which Pascal he was talking about. To convey information, papers should be formal and express things precisely. In fact, everything should be said twice, once informally for entertainment and once formally for precision.
In short, I hate when papers (especially those with one author) say we instead of I. The word we is perfectly fine to mean ``me (the author) and you (the reader)'' as in ``we can now see why...'' But there is no excuse for uses that are simply incorrect.