To be honest I am getting confused about who is arguing what here. But beliefs are central to any argument of risk compensation. People proceed on the basis of what they believe to be true. Since the level of risk is itself under debate in most relevant scenarios, and anyway none of us carries around a table of accident rates at all the locations through which we are passing on a given journey, it's impossible for us to proceed solely on the basis of facts anyway.
But any idea that risk compensation completely negates the benefits of any safety measure is ridiculous. I think I mentioned already the success of shared spaces schemes such as in
Poynton. In risk compensation terms, those work because drivers over-estimate the risks of the change, and over-compensate by taking fewer risks themselves, so the net effect is very positive. That again, though, is an example of drivers getting the assessments wrong and failing to compensate precisely. I don't see why it can't happen in reverse as well.
I believe a more typical outcome to be that safety measures produce benefits, but not as great as expected, because road users take some of the benefits as performance gains (by taking more risks). For example, the effect where a road is straightened; risk reduces, so vehicle speeds increase, and there are more crashes down the road at the next bad bend. That may well still be fewer crashes overall though. With seat belts, I'm not aware that anyone claims anything but a reduction in casualties among car occupants - although there have been claims that these are negated by increases in injuries to pedestrians and others, presumably because drivers took the benefits as performance increases and drove faster. That said, belts came in by stages, and you wouldn't expect as clear an effect from a change to require rear seat belts, because the driver is not directly affected.
In any case, getting back to helmets, my perception is that the whole thing was the other way around. I did not encounter risk compensation, rotational injuries, or any other effect as a reason not to wear helmets. I first came across them as competing attempts to explain why helmet wearing was not producing the expected level of benefits at a population level. I think the first paper I saw showing negative effects overall was in the
Journal of Products Liability in 1988, although memory could be playing tricks. Now of course, since then, there have been much more positive results as well as some more negative ones, but it's from that kind of result that risk compensation came into the debate, at least as I remember it.