These are soft sciences. You're not looking at a hard science; you can't do an fMRI and say "look at how the addiction-response area lights up when this guy is playing MMORPGs."
The Wikipedia definition does a fine job of drawing the line, I'd like to point especially to:
. . . has begun to harm the individual or others physically, mentally or socially
This question doesn't spontaneously arise from the subset of alleged addictions that you point out. How do you know somebody doesn't just like cigarettes very much? You're always drawing a line based on some criteria.
We know that a person may become addicted to video games (for instance) because there have been people who let their lives be dominated by playing video games to the extent of their family leaving them, and them losing their jobs due to high absence.
The criteria for "harming the individual" is always sketchy. Certainly, it is conceivable that one could engage in a mild form of an activity that might be harmful. Drinking reasonable amounts of alcohol, or even a bit too much occasionally, might be somewhat harmful, but we'd probably argue that it isn't an addiction if the person doesn't have a problem keeping away from it for a considerable amount of time. I'd rather place emphasis on the "or others" part of the definition. If your partner suffers because of your habit, we may be inclined to call those habits addictive. On that note, see the website GamerWidow. Of course, even this criteria has to be coupled with a lot of others. You can't jump immediately from "someone was harmed" to "therefore, I'm addicted to this causal activity."
Again, soft science. I don't think anybody can ever point to a single defining factor, or any definite threshold of anything, that either prove or disprove addiction. Think of it, rather, as a set of weighted inclusion and exclusion criteria that adds up, and that you draw a subjective line.