Risk compensation (or risk homeostasis theory) is an important explanation of behavior. For instance, many updated street designs include narrower lanes, which we assume slows drivers (as they adjust speed to compensate the new risk).

There are key limits to the theory, of course. We rarely have perfect information to make decisions, and this includes information on risk. Our risk taking behaviors (and adjustments as predicted by compensation) are based on imperfect assessments of situational risk.

That's my take anyway -- that risk compensation is a useful socio-behavioral construct, but like most such constructs, should not be used to justify individual choices.