SCIENCE@home

Theoretical and empirical analysis of the evolution of cooperation

P. Bednárik

Published 2014 in Unknown venue

ABSTRACT

Cooperative behaviour is widespread among humans and throughout the animal kingdom. Previous models suggest that the evolution of cooperation can be enhanced by network structure. However, recent experiments were not able to detect cooperation-enhancing capabilities in static networks. Only if the network is dynamic, experiments with humans report increased cooperative behaviour. Since dynamic networks imply the possibility of changing partners, an important aspect is how costs for changing partners affect behaviour. Since this aspect has been neglected so far, chapter I of this work is dedicated to close this gap and explore the effects of costs on dynamic networks. I showed that the willingness to break links is drastically reduced when links to new partners are costly. For very high costs, the rate of breaking links was so low that the network was nearly static. Interestingly, cooperative behaviour stayed at a high level nevertheless. This implies that cooperative behaviour depends, above all, on whether there is an option to switch partners or not. Even if costs are so high that this option is rarely used, cooperation levels are substantially higher than without the option. Chapter II of this thesis is dedicated to the investigation of decision-making. In the so-called Judge-Advisor-System, one person, the judge, estimates an unknown quantity. Then, the judge receives advice from another person, the advisor. Importantly, the estimates by the judge and the advisor are made independently. The task is to find out how the judge should best use the information from the advisor. Existing approaches mainly focused on two methods, (i) taking the average, and (ii) choosing one of the initial estimates. This simplistic approach is mainly driven by empirical data, where it seems that in some experiments over 70% of participants used one of these methods. However, other weights are also frequently assigned and a thorough theoretical investigation of optimal weights is necessary. Therefore, I derived a normative model that tells under which circumstances it is better to (i) take the average, to (ii) choose what you think is the better estimate or to (iii) try to assign proper weights. Which of the three is the best depends on the difference in expertise of judge and advisor as well as on the judge’s likelihood to know this difference. If the judge has a good representation of this difference, assigning weights is always the best bet. The simple average is useful if the difference in expertise is small or difficult to guess. Finally, choosing performs well if the difference is large but its amount is difficult to guess. Motivated by previous approaches, I also explored the performance of a combination of choosing and averaging, i.e. a method that uses averaging for small difference in expertise and choosing for a large difference. Surprisingly, the performance of this combined method was very poor. The main reason is the uncertainty in guessing the difference of expertise. Therefore, assigning proper weights is almost always better than using the combined method. Since choosing the worse expert has performs so poorly, the combined method requires that the risk of choosing the wrong person is low. But this means that the difference in expertise is easy to guess and therefore weighting is the best method. Over all, I showed that weighting is a viable method for a wide range of parameters.

PUBLICATION RECORD

CITATION MAP

EXTRACTION MAP

CLAIMS

  • No claims are published for this paper.

CONCEPTS

  • No concepts are published for this paper.

REFERENCES

Showing 1-78 of 78 references · Page 1 of 1

CITED BY

  • No citing papers are available for this paper.

Showing 0-0 of 0 citing papers · Page 1 of 1