The goal of this project is to build a platform for research in behavioural game theory on social networking sites which will enable experiments of unprecedented scale, resolution, interactivity and social embedding. We aim at testing the behaviour of real people in game theoretic interactions in social networks. How do people negotiate with one another? How can we aggregate opinions of individuals to arrive at high-quality decisions? In what ways do people reciprocate other people’s actions?
The ways in which people interact, socialize and communicate have dramatically changed in recent years due to internet-based technologies. This is a networked world that puts people in a closer-than-ever mesh. People regularly exchange information through internet social networks, communicate their opinions through internet forums and websites and base more and more of their economic decisions on information accessible through the internet. Many such interactions are strategic in nature: many organizations now invest in advertising and branding in social networks, and people use such networks to find jobs or business opportunities.
Classical game theory makes predictions regarding how rational agents would behave in strategic settings, such as in the domain of advertising, during business interactions, and on the job market. Although classical game theory does allow making predictions regarding human behaviour, in many domains, and especially network domains, it makes very strong assumptions. For example, the agents in classical game theory are assumed to be completely rational: they make their decisions solely based on maximising utility, they are capable of performing very complex reasoning and they assume that their adversaries are equally rational. Humans in the real-world, on the other hand are quite different. Their behaviour is sometimes emotional, they sometimes base their decisions on concepts such as fairness and reciprocity (rather than only on the monetary amount they get), and are bounded in their reasoning capabilities and often use heuristic reasoning.
The goal of the project is to test the behaviour of real people in game theoretic interactions, and especially those that take place in social networks. Some example questions are: how do people negotiate with one another? How does such negotiation take place in social networks? How can we aggregate opinions of individuals to arrive at high quality decisions? In what ways do people reciprocate other people’s actions?
Our next major goal in the project is establishing a “Facebook Game Theory Lab”. This is platform that would allow us to carry out experiments in which people interact with their friends in a set of games, designed to examine the strategic interactions that occur in a principled way. Example games are competition and resource allocation games and negotiation games. This way we can examine where peoples’ behaviour differs from predictions from classical game theory, and devise tools to better predict the behaviour.
We believe that carrying out research in behavioural game theory on Facebook offers unique opportunities.
- Through “viral marketing” mechanisms we will be able to recruit a number of subjects that is larger by orders of magnitude from previous studies.
- Larger scale will allow an as yet impossible degree of resolution with respect to key socio-demographic and behavioural variables and will broaden the population considered beyond the narrow group of psychology grad students often recruited for such studies.
- The embedding in the subjects’ social graph allows unprecedented types of analysis connecting individual variables with network characteristics and games played on the “playing field” of the social graph.
- The artificial “lab setting” of traditional experiments in behavioural game theory is replaced by the natural social habitat of the subjects, their circle of friends.
Microsoft Research - Project Waterloo
Our first project is a game played on Facebook called Project Waterloo, an example in the "Colonel Blotto" style. The game is played by two players; each has 100 troops at their disposal and needs to allocate those troops among five battle fields. Once both players have allocated their troops, the winner is determined by finding which of the two players has won more battles. A battle is won by having allocated more troops to it than the opponent.
The game is complex from a game-theoretic perspective, involves randomized strategies, and can be approached by reasoning about the opponent's reasoning. We have also found it to be fun, engaging, and slightly addictive. It is thus a great test case for studying actual strategic behaviour of people on Facebook.
Microsoft Research - Doubloon Dash
Doubloon Dash is Facebook app designed and developed by Microsoft Research to study how people participate in all-pay auctions. All-pay auctions are a particular kind of bidding mechanism in which all the participants have to pay the amount they bid no matter if they win the auction or not. Although the mechanism of all-pay auctions may sound strange, it is very common in day-to-day socio-economic situations.
One example of an all-pay auction is the job application process: All applicants make a “bid” by investing their time to prepare for and attend the interview, but only one of them receives winning utility by actually getting the job. In Doubloon Dash, players play the role of pirates who are searching for treasure. They compete with each other to reach the treasure by spending money (bids) on their ships. However, only the pirate who spends the greatest amount of money gets the fastest ship and reach the treasure first. The money and resources of all the other pirates will be lost.