In another thread, we have a small debate going on about whether to rely on a human buddy or a redundant air source to deal with OOG emergencies. I haven't seen anyone mention
Risk Compensation in discussing the question, so here is a thumbnail sketch:
The idea is that people have a certain tolerance for risk, and they will attempt to obtain as much utility from a situation as possible while keeping their risk within "acceptable" limits. Note that we are talking about their perception of risk and utility.
This means that if you "intervene" in an activity by making it safer, people will route around your intervention by acting more recklessly. The most famous example is anti-lock brakes: although cars with anti-lock brakes are statistically safer than cars without, the effect is much, much more modest than predicted by the technology. Supporters of the theory suggest that people drive more recklessly and/or drive vehicles that are harder to control given ABS to save them.
So...
Let's take a large number of diving teams of roughly equivalent experience doing roughly similar dive profiles, say PADI AOW divers with 25-75 dives each doing NDL OW dives. They will behave similarly to each other, each team will try to have as much fun as possible while staying within their perception of acceptable risk.
Now we give some teams pony bottles. We may be tempted to think that the teams with pony bottles are much safer than the teams without. And they may be safer. But the theory tells us that those teams will behave in riskier ways to try to maximize their enjoyment, perhaps by straying further away from their buddies, perhaps by checking their SPGs less often, perhaps by turning their dives much later because "they have plenty of air in reserve."
In each case they are trying to trade their additional safety for fun.
They may still be safer than the divers without pony bottles. But the theory cautions us that we should be conservative when trying to decide whether safety devices actually make us safer and if so by how much. In theory they do. In practice, they may not, or not as much as we think.
Lest you think I am arguing against safety devices, Risk Compensation makes prediction about skills as well. Let's say we take a diver and put them though GUE Fundamentals. They are now much safer on recreational dives than they were before their course.
Risk Compensation suggests they will use their new-found skills to dive in more dangerous environments like deco dives, wrecks and caves, trading the increased safety provided by their skills for increased utility. Again, the theory simply predicts that people have a tolerance for risk and will compensate for perceived safety by trying to extract more utility.