Gerd Gigerenzer has a very interesting book about this:

 http://www.amazon.com/Risk-Savvy-Make-Good-Decisions/dp/0670025658

His main point is the fact that, commonly risk and uncertainty concepts are considered to have the same meaning. Additionally, you could get some clues from "The Black Swan" by Nassim Taleb.

Oscar,

One thing you need to be aware of in looking at this is that risk compensation and risk homeostasis are different things, even though the risk homeostasis people like to mix them together to make risk homeostasis sound more generally accepted than it actually is.

Risk compensation - people changing their behaviour in response to physical and procedural mitigations - is unquestionably a real thing, and one of the reasons why mitigations are seldom as effective as they are predicted to be. I would be surprised if you find anyone arguing that it isn't real. 

Risk homeostasis - the idea that people have a natural risk "set point" and adjust their behaviour to sustain a constant level of risk - is a fringe theory. There are one or two examples that can be cherry picked to match the pattern, but the premises and conclusions are not generally true.

If you're after an alternative in the same space, I recommend looking at the idea of "revenge effects". There's a very readable book "Why Things Bite Back: Technology and the Revenge of Unintended Consequences" by Edward Tenner.

Instead of riks compensation, we often speak about behavioral adaptation in the context of traffic psychology. There is a recent book by Rudin-Brown and Jameson that my be of intereset for you:

http://www.crcpress.com/product/isbn/9781439856673

If you take a look, you will see that there is evidence that these behavioral adaptations may occur after the introduction of safety measures in traffic. (In case you`re interested, we also found evidence of behavioral adaptation caused by predictive warning systems, see attachement)

Article Behavioral adaptation caused by predictive warning systems –...

Pardon me for butting in here. I must, however, take issue with Andrew Rae's statement that risk homeostasis posits a constant level of risk. Risk homeostasis is not an outcome, let alone a constant outcome (reflecting a "natural risk set-point, a misconception widely spread by Wikipedia and related websites, as well a by others). On the contrary, risk homeostasis is a process linking outcome to a target outcome. The target level of risk is that level of risk that allows the operator (driver) to maximize the overall benefit of risk taking (time gain, economic benefit, excitement, competition, etc.)  The target varies, just as core body temperature that  is low in the early morning and high in fever, yet both low and high temperatures being homeostatically controlled. And just as the target varies, so does the outcome. The target level of accident risk is low in periods of an economic slowdown, and when operators are promised a bonus (incentive) for remaining accident-free. So risk homeostasis drives behavioural compensation (behavioural adaptation). It follows that the art of accident reduction (per capita, not per km driven) depends on the art of reducing the accepted level of risk in the driver population. The same holds for people at work. For more information you may check the website . Best regards, Gerald Wilde ([email protected]).

Two more quotes to show the misconceptions about risk homeostasis as published in Wiki websites (accessed today):

Psychology Wiki says: The theory of risk homeostasis states that an individual has an inbuilt target level of acceptable risk which does not change.

The Full Wiki states: The hypothesis of risk homeostasis holds that everyone has his or her own fixed level of acceptable risk.

Where do these notion of inbuilt and fixed come from? From hearsay? Certainly not from anything I wrote. I wished someone would correct these wiki sites, someone who  would also understand that I dislike the idea of doing it myself.

With best wishes, Gerald Wilde ([email protected]).

Gerald,

I understand the frustration in not being able to update it yourself. I agree that I oversimplified. The point I was trying to convey is that it is relatively constant - i.e., less variable than the actual risk. The misunderstanding may come because lots of the supposed evidence for risk homeostasis requires risk acceptability to be constant in order for the evidence to support the theory. If risk acceptability is taken to be variable, how do you empirically support a claim that individual behavior seeks to match that variable? Certainly not with the types of before-after studies that you used to support the theory as you proposed it.

Once again, homeostasis is a process, not an outcome, let alone an invariable outcome.

If the target value of risk is altered, so is behaviour, either in a  that is attached to behaving safely. Chapter 11 of Target Risk (3rd edition downloadable from riskhomestasis.org) gives numerous examples of effort to reduce risk tolerance in road users and how they reacted by behaving in a manner that led to fewer accidents. Ciao, Gerald