The evolution engine--random variation and natural selection--is a search algorithm for finding solutions to an adaptive problem. Given enough time, the solution is elegant: a spider's web, a hexagonally-celled beehive, or a bird's wing. Natural selection's solutions may sometimes seem of dubious elegance--the camel is as ugly and disagreeable a creature as you could find, and, while many flaunt the human eye as an example of intelligent design, what about the failure-prone human knee and back? Unfortunately, evolution does not necessarily arrive at the most universally elegant solution; it does so within the constraints of initial conditions and time: evolution is a random walk, not jump, through genetic space, and it is entirely blind. The human animal evolved from a four-footed creature; natural selection had to work with the material available. Maybe in a few million years we will somehow evolve to no longer need knee and back surgery. And the camel is indeed an elegant solution to the problem of desert survival, if not polo transportation.
Not only genes play a role in evolution: so do memes, the smallest units of cultural information, and the environment also changes. For example, say at some point some primitive humans got the idea to cook their food, and the meme spread. They would then evolve genetically to prefer it or even need it cooked, leading them to cut down more trees for firewood. Deforestation could then lead to a change in the food supply, both of animals and edible plants, which would again affect the people's diet, exerting a changing environmental pressure.
Some curmudgeons maintain that evolution does not lead anywhere, that there is no objective progress, that Time might have an arrow but Evolution does not. Some even suggest that the evolution of intelligence leads inevitably to destruction by environmental changes, which seems unreasonably pessimistic. Other people, like Robert Wright in his book Nonzero: The Logic of Human Destiny, have an almost religious sense that evolution is taking us somewhere higher. 
Evolution finds winning strategies to the game of life, whose only goal is maximum long-term reproduction. The classic example of game theory is the Prisoner's Dilemma, in which two men accused of committing a crime together are interrogated separately. They must choose between two strategies: maintain innocence (Cooperate) or confess (Defect). If they both cooperate, then they both get a medium-sized payoff (3 in the diagram) interpreted as a medium amount of time in prison. If they both defect, they both get a very small payoff (1 in the diagram), interpreted as a lot of time in prison. If one defects and one cooperates, the defector gets a big payoff and the sucker gets zero, meaning that one goes free and the other gets locked up for life. If you add the payoffs for each scenario, Cooperate-Cooperate gives the greatest total (6), and Defect-Defect gives the least (2). Obviously, for the greater good the two players should be nice and cooperate. Problem is, logically they shouldn't: since neither player knows what the other will do, and each is trying to maximize his own possible winnings and minimize his own possible losses, they perversely choose Defect-Defect.
The problem with that logical scenario is that it doesn't match real life: most people in an analogous situation would be nice and choose to cooperate. Most games are not played just once, but many times. If you were to be caught in a repeated Prisoner's Dilemma, you would probably cooperate the first time, and then base your strategy in subsequent rounds on what the other player did in the past. If you were nice and he betrayed you, you would certainly retaliate, and probably enjoy it. Later, if the other player began playing nicely, you might choose to forgive (but not forget) his past mistake. If the other player turned out to be incorrigible, always choosing a nasty strategy, you would not only be nasty back, but probably try to avoid the nasty player and find someone nicer to play with, leaving the nasty players to defect on each other. Ideally, the reputation of each player would be publicly available, so that you would have an advance notion of a player's tendencies even if you had never encountered one another.
Evolutionary game theory predicts just such scenarios, and Robert Axelrod's clever computerized tournaments have shown that cooperative strategies tend to beat out nasty strategies over time. Mutations and disruptions in the environment can temporarily re-introduce nasty strategies, so game-players may want to have a retaliatory strategy in their back pocket, but they should also have the ability to forgive and return to happy, cooperative equilibrium.
The actual participants in the game of life are not human beings or other organisms, but genes and memes, which are just two (important) examples of replicators. If a gene wins a round, an organism carries it into the next generation. Sometimes a gene is competing against the environment (e.g. a gene for dark skin might be adaptive near the equator but maladaptive in the colder north), which is not very interesting, but sometimes it is competing with other genes, which is where game theory comes in. Repeating a game endlessly will evolve a set of genes coding for the most successful strategies. For example, human beings have genes providing several strategies for dealing with other people, e.g. kill, steal from, help, trade with, beg from, etc. Which of those strategies a person selects will depend on which are open to him (a weak person may not have the option of a violent strategy) and the environment (a stable, peaceful environment with many cooperators is more open to cooperative strategies, whereas an uncertain environment with a lot of nasty defectors lurking about favors negative, or at least risk-reducing, strategies). In the case of human beings, it may be that strategies come in sets, e.g. generally cooperative, distrustful, exploitative, or violent, and that our childhood environment leads us to "choose" a long-term, even lifelong, strategy set. From this would come the common understanding of someone's character being good (inclined to cooperate) or bad (inclined to defect).
Niels Bohr once said that if thinking about quantum mechanics does not make you dizzy, then you have not really understood it. The same can be said about replicators. It is not human beings that replicate, but genes and memes that do; we are simply the medium, the carriers.
Pay no attention to that man behind the curtain.