We can see this principle at work in the fines for things such as littering, illegal parking, and drunk driving. Individual instances of these crimes gain a person a small amount of value; the fines for being caught are around one or two orders of magnitude greater than this (at first--often repeat offenses raise the expense another order of magnitude in the form of either jail time or increased insurance costs), in line with the approximation of being caught a non-negligible fraction of the time.
By comparison, illegal file sharing has a vanishingly small chance of getting your life completely destroyed. The problem with this is two fold. First, it isn't at all calculable what the expected value is; there simply isn't enough data to accurately determine what the probability of getting caught from any single illegal act is. Second, the enforcement rate is sufficiently low that, even if it were reliably calculable, people would not emotionally treat it correctly--the same emotional detachment from strongly negative consequences occurs here as the one that makes people take other sorts of extreme risks such as teenage unprotected sex, extreme sports, and going to war as a foot soldier (note that I am not saying that these always have negative expected value; merely that people are not good at estimating these sorts of risks).
The problem the RIAA faces is that enforcement to the degree necessary for a lower level fine to create sufficiently negative expected value is a practical impossibility; there's no reasonable way to effectively catch and fine even 1% of file-sharers (a rough estimate of what order of magnitude of enforcement is necessary to make people effectively able to gauge the risk--such an enforcement rate would not stop illegal file sharing altogether, but it would at least slow it down, much like my other examples). The only long-term option is to change business models to one less completely invalidated by the advance of technology.