When we must choose between cooperating with others or betraying them, we are more likely to cooperate if they have acted like this with us or if we ourselves have behaved altruistically. We do so because learning reinforces what has gone well for us and not because we imitate successful people. These are the conclusions highlighted in a study conducted by the Carlos III University of Madrid based on the prisoner’s dilemma, a popular model for studying conflict.
The protagonist of Deux jours, une nuit, the latest film by the Dardenne brothers which premiered at the Cannes Film Festival, must convince her colleagues to give up their bonuses so that she can keep her job. This is a specific case of the prisoner’s dilemma, a formula commonly used in game theory to analyse human cooperation.
The dilemma in question presents a scenario where the police arrests two people suspected of committing a crime and isolates them. If they both cover for one another or remain silent then they will both be sentenced to a year in prison. If only one of the suspects betrays the other then the betrayer walks free while the other suspect will spend three years in jail, but if both suspects betray the other then they will both be sentenced to two years.
“It is a paradigmatic approach where the best thing we can do is cooperate and things will go well for both of us,” Anxo Sánchez, professor of the Interdisciplinary Complex Systems Group at the Carlos III University of Madrid (UC3M) commented to SINC. “Although, if you cooperate and I betray you, things will be even better for me but at your expense, of course; therefore, if you think the same, you should betray too, but in that case things will go badly for both of us, and quite a lot worse than if we had cooperated. In these circumstances, which we are often faced with in our lives, it is hard for cooperation to emerge, and yet this does occur”.
Based on an analysis of the decisions made by the participants in this game, the researcher and other colleagues in Madrid and Zaragoza have shown that when we have to choose between cooperating with others or betraying them, we are more likely to cooperate if people have previously cooperated with us, as well as in cases in the past when we personally have acted in this altruistic manner.
“In other words, when you betray once, it is quite hard for you to cooperate in the future, whereas if you have already cooperated then you will have a tendency to continue doing so,” Sánchez clarified and underlined that the experimental evidence confirms this tendency.
In order to explain this phenomenon, Sánchez and the researcher Giulio Cimini from La Sapienza University in Rome have developed a model which evaluates the different forms in which people adapt their behaviour. The results, published in the ‘Journal of the Royal Society Interface’, reveal that only what is known as ‘reinforcement learning’ can explain the behaviour and the level of cooperation observed.
“Reinforcement learning is a process whereby if we do something and we believe that it has gone well, we will be more likely to repeat this action; whereas if we think that it has gone badly, we will be less likely to do so,” Sánchez pointed out.
This might seem to be common sense, but it contradicts established approaches: “Up to now the majority of models on the origin of behaviour were more social, that is, more associated with imitating others, either because we know them or because we think what they are doing is good, but our work definitively rejects this,” Sánchez said, and went on: “Our behaviour changes according to how things have gone for us, not because we are copying others socially”.
The work also rejects some models used by economists: “In principle, they assume that we see what others have done and calculate how to behave so that, if others repeat their actions, we get the best possible outcome. Yet this does not fit in with the experimental findings, in other words, we are not that calculating. Therefore, many cooperation-related economic results obtained theoretically based on the supposition that we are perfectly rational are no longer valid”.
Recommendations for your personal life
According to the authors, the conclusions can also be applied to our personal and professional lives. On the one hand, they recommend not falling into a situation of desperation in which you only betray, because this will not contribute to maintaining or recovering cooperation. According to Sánchez, “you have to be more willing to cooperate after having betrayed in order to see if others follow suit”.
On the other hand, the study shows that the most successful behavioural strategies are those with an intermediate level of aspiration. In other words, we must aspire to obtain reasonable gains from our interactions: not too high as this leads us to betray and then to us being betrayed; yet not too low either, because in this case we would be exploited all the time.
Giulio Cimini y Ángel Sánchez. “Learning dynamics explains human behaviour in Prisoner’s Dilemma on networks”. Journal of the Royal Society Interface, 2014.
For media only:
If you are a journalist and would like to contact the researchers, please register as a journalist in SINC.