I am not satisfied with most definitions of science I've seen so far, either essentialist or enumerative. I think the "social" definition below is much better:
1. Science is what scientists do .
It may not be very satisfying, but neither is it as glib as it first appears. Still, I believe it needs some precisions: obviously, not everything scientists do is science. This applies in private life (some scientists are atheists and others are deeply religious) but also to public attitudes: Johannes Stark was an authentic Nazi, but this does not prevent us from studying the Stark effect. Conversely, his scientific achievements do not render Nazism respectable. Then:
2. Science is what scientists do in their professional capacity.
This works better, but without defining the "professional capacity" the definition is circular. If we stick with the social criterion, what matters is not what scientists do by themselves at work, but rather how they interact professionally. Let us try:
3. Science is what scientists communicate to each other in community-accepted venues.
where the venues (taken in the widest sense of the word) vary between scientific domains. In the "hard sciences", they are mostly peer-reviewed journals and conferences, but may also be informal reunions that mix several conversation topics. The latter case poses no problem: it is still quite clear to the participants whether they are discussing science, administration or simply gossiping.
Even papers published in the most respectable journals sometimes contain claims (generally indirect) that are not "adequately supported by the experimental data", without negative consequences (except maybe editorial dyspepsia). For instance, many works on nanoparticles mention their potential therapeutic applications without any attempt at... This is just introductory (or concluding) fluff and is recognized as such by the scientific audience, though not always by science journalists (often leading to sensationalistic titles in the general press and on the Internet.) I would therefore ammend version 3. of the definition to:
4. Science is what scientists communicate to each other in community-accepted venues, as long as it is taken seriously by other scientists.Of course, "taken seriously" need not mean "accepted": errors are taken very seriously, while fantastic projections ("our discovery will be in mass production within five years and will completely change the world within ten") are not.
 Percy W. Bridgman, "On Scientific Method" in Reflections of a Physicist (1955).↩