Thomas C. Schelling, Micromotives and Macrobehavior, New York, W.W. Norton & Company, 1978.
When we analyze how people behave in trying to escape from a burning building we mean that they really are trying to escape. They are not simply acting, 'as if' they dislike being burnt. 
How well each does for himself in adapting to his social environment is not the same thing as how satisfactory a social environment they collectively create for themselves. 
The body of a hanged man is in equilibrium when it finally stops swinging, but nobody is going to insist that the man is all right.
Calling it an equilibrium does not imply that everybody - or even anybody - likes the seating arrangement, only that nobody alone can do better by changing to any available seat. Nor does it imply that there are not alternative seating patterns, very different ones, that could also be equilibria. 
Now for what is special about economics. Economics is mainly concerned with transactions in which everybody affected is a voluntary participant. 
If you buy somebody else's book, I may feel 'affected' by the transaction because the alternative I had in mind was selling you my book instead. I can wish that people wanted, and would pay for, the things I have to offer, and would offer me, at attractive prices, the things I would like to buy; but this is more like wishing for transactions that didn't occur than objecting to some that did. 
[marketplace's voluntary dynamic is highly efficient for production of conscientiousness and material innovation]
The free market, when it works, is that special case of knowledgeable voluntary exchange of alienable commodities. 
It may or may not occur to me that I am part of your problem as you are part of mine, that my reaction to the environment is part of the environment, or that the quantity or number I am responding to is the sum of the reactions of other people reacting like me. 
George Akerlof argued that the seller of a used car konws whether or not it is a lemon; the buyer has to play the averages, knowing only that some cars are lemons but not whether the particular car he's buying is. Buyers will pay only a price that reflects the average frequency of lemons in the used-car crop. That average is a high price for a lemon but understates the worth of the better cars offered on the market. The owners of the better cars are reluctant to sell at a price that makes allowance for the lemons that other people are selling; so the better cars appear less frequently on the market and the average frequency of lemons increases. As customers learn this, they make a greater allowance for lemons in the price they're willing to pay. The cars of average quality in the previous market are non undervalued and their owners less willing to sell them. The percentage frequency of lemons continues to rise. In the end, the market may disappear, although institutional arrangements like guarantees, or the certification of cars by dealers who exploit a reputation for good cars, may keep the used-car market alive.
Akerlof generalized this model to a number of markets in which there is unequal information on the two sides. 
[critical mass dilemma: market disappearance]
If it is proportions that matter - smoking cigarettes or wearing turtlenecks or speaking with a particular accent depending on the fraction of the relevant population that does so - there is the possibility of dividing or separating populations. If people are influenced by local populations - the people they live with or work with or play with or eat with or go to school with or ride the bus with, or with whom they share a hospital ward or a prison cell block, any local concentration of the people most likely to display the behavior will enhance the likelihood that at least in that locality, the activity will reach critical mass. 
[reorganization of network localities can induce critical mass mechanics]
Once travelers got used to Garrett Morgan's traffic lights, they learned that it was dangerous to cross against a flow of traffic that was proceeding with confidence. The lights created the kind of order in which non-compliance carried its own penalty. And there was impartial justice in the way the lights worked: unable to recognize individual travelers, the lights could hurt no one's feelings by not granting favoritism. 
[self-enforcing conventions require induced penalty for non-compliance and impartiality of coordination mechanism]
People do things, or abstain from doing things, that affect others, beneficially or adversely. Without appropriate organization, the results may be pretty unsatisfactory. 'Human nature' is easily blamed; but, accepting that most people are more concerned with their own affairs than with the affairs of others, and more aware of their own concerns than of the concerns of others, we may find human nature less pertinent than social organization. These problems often do have solutions. The solutions depend on some kind of social organization, whether that organization is contrived or spontaneous, permanent or ad hoc, voluntary or disciplined. 
A good part of social organization - of what we call society - consists of institutional arrangements to overcome these divergences between perceived individual interest and some larger collective bargain. 
Within the family we can save hot water on Friday night by taking brief showers, rather than racing to be first in the shower and use it all up. But that may be because within the family we care about each other, or have to pretend we do, or can watch each other and have to account for the time we stand enjoying the hot water. It is harder to care about, or to be brought to account by, the people who can wash their cars more effectively if I let my lawn burn, or who can keep their lawns greener if I leave my car dirty.
What we need in these circumstances is an enforceable social contract. I'll cooperate if you and everybody else will. I'm better off if we all cooperate than if we go our separate ways. In matter of great virtue and symbolism, especially in emergencies, we can become imbued with a sense of solidarity and abide by a golden rule. We identify with the group, and we act as we hope everybody will act. We enjoy rising to the occasion, rewarded by a sense of virtue and community. And indeed a good deal of social ethics is concerned with rules of behavior that are collectively rewarding if collectively obeyed (even though the individual may not benefit from his own participation). But if there is nothing heroic in the occasion; if what is required is a protracted nuisance; if one feels no particular community with great numbers of people who have nothing in common but connected water pipes; if one must continually decide what air-conditioned temperature to allow himself in his own bedroom, or whether to go outdoors and check the faucet once again; and especially if one suspects that large numbers of people just are not playing the game - most people may cooperate only halfheartedly, and many not at all. 
The trouble is often in making the bargain stick. 
Whatever the technology of cooperative action - whether every litter bit hurts, or the first few bits just about spoil everything - people who are willing to do their part as long as everybody else does, living by a commonly shared golden rule, enjoying perhaps the sheer participation in a common preference for selflessness, may have a limited tolerance to the evidence or to the mere suspicion that others are cheating on the social contract, bending the golden rule, making fools of those who carefully minimize the detergent they send into the local river or who carry away the leaves they could so easily have burned.
There are the cases, though, in which not everybody gains under the social contract. Some gain more than others, and some not enough to compensate for what they give up. 
[emotional reactions maintain the balance between cooperative and individual advantages - charity to spite]
The prisoner's dilemma generates an inefficient equilibrium: There is one way that everyone can act so that everybody is doing what is in his own best interest given what everybody else is doing, yet all could be better off if they all made opposite choices. This calls for some effort at social organization, some way to collectivize the choice or to strike an enforceable bargain or otherwise to restructure incentives so that people will do the opposite of what they naturally would have done.