The major problem with overt incentives linked to a single measure is, as your examples illustrate, the resulting focus on hitting a particular target. In general, people are simply too stupid to design adequate incentives, which is why in general open markets deal better with resource allocation than command & control economies. The other problem with targets is again human stupidity. Our tiny ape-brains are rubbish at imagining a different future; we simply extrapolate in a straight line from today. Your dentist example is an excellent illustration of this: hypothetically, dentists will see reduced income if they encourage measures that improve dental hygiene. In reality they increase their income. The reason this occurs is because once people aren’t losing their teeth to decay, they aspire to perfection. UK dentists face each day knowing it will be extractions and root canals; Brits have terrible dental hygiene. US dentists on the other hand rarely perform extractions or root canal surgeries; they are making a fortune straightening imperfections, applying caps and crowns, enamels, and other cosmetic treatments. Once one need is satisfied (keeping one’s teeth) new needs appear. Static targets fail entirely to take this reality into account because they imagine a static world.
Your example of attempts to reduce plastic waste by charging for supermarket shopping bags is another excellent illustration of generic human stupidity. While people are using fewer plastic bags, the supermarkets are busy wrapping absolutely everything in plastic. It’s impossible in the UK to by any vegetable, meat product, or other consumable without it being needlessly wrapped in enormous quantities of plastic. Instead of the supermarket target being “reduce plastic” it was “reduce use of plastic bags” which entirely missed the point. We’d have to stop using a billion plastic bags a day to offset the absurd levels of plastic wrapping used in every UK supermarket today.
So yes, targets are generally ill-advised because people are too stupid to set adequate targets than account for the big picture. But assuming people “generally do the right thing” is also wrong because nobody has a clue what “the right thing” is until they’re told. Fifty years ago everyone smoked and drank far too much alcohol; today that problem is slightly reduced because of increases in the price of harmful products and half a century of public education. Today Chinese consumers are directly responsible for the extinction of a wide range of megafauna including rhino, elephant, tiger, and pangolin. In their minds they are doing “the right thing” by signaling to their friends and family how wealthy they are: they can afford to purchase items made from the body-parts of these endangered animals. But of course, looking at the larger picture, they are driving all the animals into extinction, which is stupid and cruel and pointless. This is, incidentally, a classic example of market failure: there’s a clear market and a clear buyer-seller relationship, but the consequence of this particular market is catastrophic. As we humans have zero capacity for modifying our actions (immediate gratification always wins over longer-term interests) it’s clear that if we want to ameliorate our harmful impact on the world around us, we need a great deal more than hoping that people will know “to do the right thing.”