Ivory Towers: Computer simulation and the con man

CONFIDENCE tricksters are the theme of a beguiling piece of sociological game theory, The Evolution of the 'Con Artist', by Lee Alan Dugatkin (Ethology and Sociobiology, Vol 13, 1992). It starts with a problem in ethics known as the Prisoner's Dilemma.

Two prisoners are accused of a crime. If both confess, they will each be jailed for 10 years; if neither confesses, they will both be released after a few months; but if one confesses and the other does not, the confessor is set free and the other sentenced to 15 years. The prisoners may not communicate with one another. What should they do?

From a game theoretic viewpoint, the best strategy is to confess, though both would be better off if neither did so. This generalises to situations where one may choose to co-operate with a partner, or to pursue self-interest by cheating. 'From an evolutionary perspective', writes Dugatkin, 'co-operation remains something of a mystery. We can all recant scenarios in which cheating yields a higher payoff than co-operating.'

To investigate whether cheats prosper, he considers a simple two-person game in which each may choose to co- operate or defect. If both co-operate they share the rewards; if one co-operates and the other defects, one makes a killing and the other gets the 'sucker's payoff'. If both defect, they each receive a pittance.

For a single game, the best strategy (depending on the size of the rewards) can still be to defect, but for a long series, a strategy known as TFT (tit-for- tat) is recommended: start by co-operating, then do whatever your partner did in the previous game. Such a strategy is described as 'nice' (it starts by co-operating), 'retaliatory' (punishes cheats by cheating back) and 'forgiving' (rewarding the reformed cheat by co-operating).

This has been used to explain the evolution of altruism, since co-operation is rewarded, but the theoretical model relies on a series of games between the same partners. What happens in the case of a con man, who always defects (the ALL-D mutant in game parlance), picking his victims from a large population?

For a computer simulation of the game, Dugatkin divides his large population into smaller patches. In any patch, the ALL-D mutant, playing against TFT-strategists, will start by doing well, but his rewards decrease as his behaviour becomes known. His best strategy is then to move on to another patch. 'Con Artists move from place to place and leave only once their cover is blown'.

But the optimal decision, according to the computer simulation, may be for a Con Artist to settle down in his own patch. Suspicion of outsiders may be a useful strategy for honest co- operators, but under certain circumstances, any patch may be ripe for exploitation.

'If conditions allow invasion', Dugatkin concludes, 'Con Artists usually equilibrate as the majority strategy.' In other words, most of us end up cheating. If you can fool all the people all the time, it is the best strategy of all.