The policy implications are more nuanced than he says. We should do one of two things: 1.) give the poor so much money that they get into the region with an upward sloping utility function, assuming such a region exists. 2.) Tax the poor, since taking some of their money will not make them worse off (what's one more bee sting?).
Mysteriously, he never mentions 2).
2.) Utility functions are mathematical tools, and our interpretations of and use for them don't exist in a moral vacuum. Why should tax policy seek to minimize it's effect on our utility functions at all? Why not take $40000 from Warren Buffet AND the primary school teacher? Local movement on one's utility function is important, but so is absolute position, with all of the attendant moral implications.
2) Tax policy should minimize impact on our utility because the goal is (in principle) to harm people as little as possible. This is why we consider it fair to take $1 million from Buffet but nothing from a poor person - the loss of $1 million will harm Buffet less than the loss of $50 would hurt a poor person.
If Karelis is right, taking $50 from a poor person won't hurt them at all - what's one more bee sting?
Sure, adding one more impossible task on top of everything else a person who views the world as a set of impossible challenges does not significantly change that person's miserable life. (Your odds of collecting are admittedly low though.)
However there are plenty of other people with similar incomes who manage to make ends meet. For those people, adding the additional burden can easily make all the difference between holding everything together and having their lives fall apart. That's a huge change in utility. (I can name two of these off of the top of my head.)
Furthermore there are other people who are temporarily impoverished. They weren't so. Now they are so. They may be better soon. Punitive measures like the ones you suggest can greatly lessen their odds of coming out of poverty. (I can easily name multiple friends of mine who were temporarily homeless with no resources, who managed to reconstruct their lives.)
There are two excellent reasons why the bee sting theory of poverty can be correct, which would invalidate your conclusion that it is not bad to heavily tax the poor.
There's no point in <i>curing</i> anything less than all the bee stings, because of the huge drop-off in their utility function. Would it make any sense at all for you to attend 1/2 of medical school if you wouldn't eventually enjoy the pay-off of becoming a doctor? Would doing half the monumental amount of work to start a successful business make you rich? And yet you probably wouldn't think that you have a flat utility function, would you?
2.) This is only relevant if you make the flat-utility function assumption, which I don't think is true.
If Karelis asserts there is no utility gain in going from 5 to 4 bee stings, then Gain = 0 = U(4 bee stings) - U(5 bee stings), or U(4 bee stings) = U(5 bee stings). That's a flat utility function.
Would it make any sense at all for you to attend 1/2 of medical school if you wouldn't eventually enjoy the pay-off of becoming a doctor? Would doing half the monumental amount of work to start a successful business make you rich? And yet you probably wouldn't think that you have a flat utility function, would you?
You've just described exactly what a flat utility function means. U(0 years of med school) = U(half of med school). And similarly, there would be no harm in kicking people out of med school who would not finish anyway.
2) I suppose this might be seen as biting the bullet and admitting to an indefensible point, but I do not find it hard to imagine that if I were poor and desperate, and some asshole came and took my last $50, and I had absolutely nothing, nothing else with no hope of being able to improve things, I wouldn't care. I was already fucked anyway.
Karelis's viewpoint is that they wouldn't have spent the $50 to alleviate the bee sting. They would have spent it on something that did give them utility. That nice thing (that they really couldn't afford!) what is taken away with the extra taxation.
That said, I think the point you make is reflected in a lot of real public policy. Poor areas get the worse public services, the most pollution, etc. They have so many bee stings that they can't be bothered to complain about one more.