Justin Li

Evolution of Morals


There was a feature in NYTimes Magazine written by the famous psychologist Steven Pinker, titled "The Moral Instinct". It talks about how moral choices are difficult to understand from psychological and philosophical points of view, and how morals are not necessarily rational.

Some where in the article is this little passage:

In his classic 1971 article, Trivers, the biologist, showed how natural selection could push in the direction of true selflessness. The emergence of tit-for-tat reciprocity, which lets organisms trade favors without being cheated, is just a first step. A favor-giver not only has to avoid blatant cheaters (those who would accept a favor but not return it) but also prefer generous reciprocators (those who return the biggest favor they can afford) over stingy ones (those who return the smallest favor they can get away with). Since it’s good to be chosen as a recipient of favors, a competition arises to be the most generous partner around. More accurately, a competition arises to appear to be the most generous partner around, since the favor-giver can’t literally read minds or see into the future. A reputation for fairness and generosity becomes an asset.

Now this just sets up a competition for potential beneficiaries to inflate their reputations without making the sacrifices to back them up. But it also pressures the favor-giver to develop ever-more-sensitive radar to distinguish the genuinely generous partners from the hypocrites. This arms race will eventually reach a logical conclusion. The most effective way to seem generous and fair, under harsh scrutiny, is to be generous and fair. In the long run, then, reputation can be secured only by commitment. At least some agents evolve to be genuinely high-minded and self-sacrificing — they are moral not because of what it brings them but because that’s the kind of people they are.

Of course, a theory that predicted that everyone always sacrificed themselves for another’s good would be as preposterous as a theory that predicted that no one ever did. Alongside the niches for saints there are niches for more grudging reciprocators, who attract fewer and poorer partners but don’t make the sacrifices necessary for a sterling reputation. And both may coexist with outright cheaters, who exploit the unwary in one-shot encounters. An ecosystem of niches, each with a distinct strategy, can evolve when the payoff of each strategy depends on how many players are playing the other strategies. The human social environment does have its share of generous, grudging and crooked characters, and the genetic variation in personality seems to bear the fingerprints of this evolutionary process.

The thing that fascinates me is how seemingly mathematical this description is. The appearance of generosity, the actual generosity, the percentage of lying, all seems to be describable through numbers, and individuals are just a function with these numbers as input. These individual agents would interact and learn, eventually leave offsprings and die. A multi-agent simulation seems to go well with what Trivers describes.

One problem that I have trying to understand this simulation conceptually is what the default values for the offspring should be. Intuitively, the offspring would be influenced by his parents (learning by example in society). There might also be some influence from the average of societal values, as the culture that the offspring grows up in. Neither of these ideas, however, give a precise mathematical definition of how to mix the two values.

I really want to try and write this simulation... when I'm not swamped with work.

Could a simulation of this be done?