I'll argue for two separate bright-line rules: the first based on individual rights (but pretty influenced by economics), and the second where I ditch the rights-based argument and stick to the economics.
Let's compare Roberts' first two examples, sticking to a rights-based approach for now; the pure economics version will be a separate post.
The only mechanism by which somebody else's obesity really hurts me is when either my insurer is legally barred from charging actuarially fair premiums, or when health care is covered through generalized taxes.* Let insurers charge the obese for their actual expected risk and there is no cost on anybody else. Let the government get out of the business of providing health care and instead get into the business of giving poor people money and mandating that they buy private health insurance with it, and obesity again puts no cost on anybody else. It's perverse to tell people:
"In exchange for forcing you to be part of a public insurance scheme that you might not want, I'm going to set up an apparatus of regulations, subsidies and taxes to force you to behave in ways that reduce the costs on the system that I'm forcing you to join. You're welcome."Roberts' long causal chain runs directly through deliberate policy choices that work to socialise the costs of individual behaviour. And if we let things run through that mechanism, there is absolutely no end to the range of behaviours that can legitimately be subject to the heavy hand of the state. Remember that STDs cost the government money too.
By contrast, carbon emissions do affect others directly and in a way that's awfully hard to handle through tort. Setting up a clean carbon tax seems the best way of minimizing aggregate rights violations: the tax is an imposition, but so too are the emissions.
When we ditch the rights-talk and stick to economics, we have to look at pecuniary, non-pecuniary, and fiscal externalities. That's likely worth a separate post since it comes up an awful lot and I'd rather like to be able to point to it on its own in future. In short, the economic case for intervening in response to external effects depends not on whether other people are bearing costs but rather on whether there are efficiency consequences.
* There can be a case where the insurer is allowed to do it but it isn't worth the insurer's time and effort. Aggregate losses in this case can't be very large - the insurer doesn't spend much time assessing your risk of paper-cuts for similar reasons. The same holds for other kinds of transactions where there are apparent effects. Airlines have tried to require that very large people purchase two seats to accommodate their girth; some countries have socialised this cost by mandating that airlines give fat people two seats for the price of one; others have banned it as discriminatory. The obese cost you money in higher airfares because the government has forced the airline to make it so.