Experiments on the web

This term I finally set some time aside to materialise the interest I have in programming in an experiment on the web. The objective was to explore the potential of conducting experiments online including ‘complex’ interactions with the participants, such as the introduction of games requiring instant calculations such as ‘rock-paper-scissors’ or the provision of feedback from previous rounds. For an example please click here.

The project started with the coding of the experiment ‘from scratch’ on PHP and the associated revision of the data stored on MySQL. Once I finished creating it, I proceeded to ask my 96 first-year students to participate in preparation for a class. It was quite rewarding to find out that all worked as planned and 59 participants took part in this pilot.

Encouraged by the results I then made a few modifications to the code in an effort to make it more efficient and useful on smartphones. Consequently, I then asked to 12 of my second-year students to participate all together at the same time in the experiment on their smartphones. It all worked smoothly! This enables me to set new targets for the next term; that is, integrate new experiments with our recruitment system and conduct proper sessions with the right incentives.

Although the results from the pilot are merely anecdotal, it is always good to find out that they match the predicted behaviour and some might offer new routes for research. In particular, the above-mentioned experiment was on cooperation with heterogeneous endowments and consisted of three stages and two treatments: “Your” and “Another”.

In the first stage, subjects were asked to play 9 rounds of ‘rock-paper-scissors’ and for each time they beat the computer they received one chocolate. In the second stage they were matched with another participant and had to decide whether to send their earnings to this person or not (depending on the treatment this person was from another group or the same group they were).

If anyone decided to send their earnings the amount was doubled. In the third stage the participants only needed to flip a coin twice to earn more chocolates.


There were no treatment differences as about half of the participants on both samples decided to send their endowments. This might be because most of the first-year participants did not have close ties with their colleagues in the group. Moreover, as can be seen in the adjacent Figure the decision to send chocolates was clearly influenced by the amount of chocolates earned in the first stage.


Finally, results from the third stage reveal that the distribution of ‘heads’ reported is biased towards higher numbers (35% in two vs. 25% predicted; 54% in one vs. 50% predicted and 11% in zero vs. 25% predicted).  This clearly might show that participants indeed decided to cheat in this part in an effort to gain more chocolates or compensate for their decision in the second stage.

In retrospect, I think that efforts of this kind should be performed more regularly in Experimental Economics. I am looking forward to the next term when I hope I can conduct new experiments. I will keep you posted!


Social Structure Summer School

From the 3rd to the 8th of September 2012, I had the opportunity to attend the Social Structure Summer School at the Courant Research Centre “Evolution of Social Behavior”, University of Göttingen, Germany.

In this week-long workshop, Ph.D. students from a wide diversity of fields (such as Biology, Anthropology, Mathematics, Economics and Ethology) had the chance to learn, discuss and integrate their theoretical, observational and experimental approaches for the study of sociality and cooperation.

During the time there, I learned new techniques for the analysis of cooperation that can definitely benefit my own research. Among those, agent-based modelling stood as the most interesting to me.


By receiving training in the use of NetLogo, we were able to create simple models capable of simulating common situations involving cooperation dilemmas. In my case, I worked on a model which tests the survival of strategies in typical coordination-failure games.

For example, the Figure attached shows a starting population of subjects with the following strategies: ‘cooperate’, i.e. always cooperate; ‘defect’, i.e. always defect; ‘tit-for-tat’ and ‘random’. After some iterations, the subjects with the strategy ‘defect’ are the only survivors. This is a common result in the empirical literature.

If you’d like to modify the model’s parameters to test your own hypotheses please go here.

In retrospect, I was glad to meet and interact with Professors and other students who study the evolution of cooperation and are not related to Experimental Economics. There is so much to learn from other approaches! Not surprisingly, the potential for future collaborations is appealing.

SSSS 2012