OK, here's my response:
There are three main uses, one scientific and the others merely traditional:
Scientific definition: the learned and shared memes (symbols, behavioral habits, and norms for creating physical objects) shared among a particular population.
Traditional definition #1: the arts and letters (origin: upper-class snobbery, see Raymond Williams, _Keywords_)
Traditional definition #2: what ordinary people do out side the spheres of paid labor and politics (origin: late 20th-century subconscious adaptation of traditional definition #1 in attempt to describe and analyze off-the-job life, in response to rise of TV and movies as dominant leisure-time activities)
There are problems, of ascending scale, with each definition:
Scientific: the concept is highly abstract and tautologous; it lacks explanatory power; we want to know what gave rise to particular cultures, not just that they exist (this latter was news to Europeans up until pretty recently).
Trad #1: obviously unscientific definition, useless for social science/rational description of human societies
Trad #2: same flaw as Trad #1; confuses issues; the worst of both worlds -- the vagueness/tautologousness of the scientific definition, plus the imprecision of Trad #1; virtually every time you see "culture" in modern social science (perhaps excluding anthropology), it really means "off-the-job practices"; by definition, economic life and politics are part and parcel of culture, if you define that term scientifically.
My nomination for the single most harmful concept in modern social science/social criticism = "consumer culture," which is a biased insult ("consumer") attached to a non-scientific muddle.
Herr Goebbels was correct about what to do when you hear the word culture, I MHO. Social science and left politics would both be much improved if the word "culture" was dropped entirely.