The cost of early selfishness is greater than the cost of trust

It seems that density and trade are likely progenitors of Trust.

From The evolution of generosity in The Economist:

Studying human evolution directly is obviously impossible. The generation times are far too long. But it is possible to isolate features of interest and examine how they evolve in computer simulations. To this end Dr Delton and Dr Krasnow designed software agents that were able to meet up and interact in a computer’s processor.

The agents’ interactions mimicked those of economic games in the real world, though the currency was arbitrary “fitness units” rather than dollars. This meant that agents which successfully collaborated built up fitness over the period of their collaboration. Those that cheated on the first encounter got a one-off allocation of fitness, but would never be trusted in the future. Each agent had an inbuilt and heritable level of trustworthiness (ie, the likelihood that it would cheat at the first opportunity) and, in each encounter it had, it was assigned a level of likelihood (detectable by the other agent) that it would be back for further interactions.

After a certain amount of time the agents reproduced in proportion to their accumulated fitness; the old generation died, and the young took over. The process was then repeated for 10,000 generations (equivalent to about 200,000 years of human history, or the entire period for which Homo sapiens has existed), to see what level of collaboration would emerge.

The upshot was that, as the researchers predicted, generosity pays—or, rather, the cost of early selfishness is greater than the cost of trust. This is because the likelihood that an encounter will be one-off, and thus worth cheating on, is just that: a likelihood, rather than a certainty. This fact was reflected in the way the likelihood values were created in the model. They were drawn from a probability distribution, so the actual future encounter rate was only indicated, not precisely determined by them.