Behavior and Rationality

Coordinator: Mario Gilli (Economics, UNIMIB)

Research promoters:
Nicola Gatti (Computer Science, Politecnico Milano),
Mario Gilli (Economics, UNIMIB), Roberto Lucchetti (Mathematics, Politecnico Milano)


Objectives: 
The object of this research program is the theoretical and systematic empirical analysis of the way agents behave and, especially, of how the combination of agents’ behavior affects social outcomes. This second point is particularly important because we are not interested in personal behavior in itself, but in understanding the hidden individual determinants of social patterns. To this aim, we believe that a unifying, but not exclusive, language is provided by game theory. Hence, this research program will partially focus on game theory, but we search also for contributions on agents’ behavior that arrive from different disciplines and perspectives.

The starting point of this program is that research on individual behavior should focus on improving the ability of current theories to improve predictions of actual behavior in a wider range of field and lab settings. To be clear, current theory is successful, because it works empirically in many important circumstances. To be equally clear: there are also circumstances in which it does a poor job empirically. The circumstances under which current theoretic concepts do a poor job can often be understood by the theory itself: when equilibria are not robust, the environment is complex, or when circumstances are unfamiliar, standard theory is less likely to perform well. In these circumstances, there is greater scope for behavioral factors to improve the available theories.

In thinking about where the analysis should go next, we plan to examine which of these forces are likely to be important, and the characteristics of a good theory that would explain them. Instead of trying to survey all of the promising avenues for improving the theory, we are planning focus on:

  1. models of complexity and of bounded rationality;
  2. economic reasoning and artificial intelligence;
  3. structural models of strategic behavior;
  4. learning processes and equilibrium concepts;
  5. models of social preferences;
  6. cognition and behavior in stochastic games;
  7. applications.

We think that these are promising routes for improving and widening our understanding of individual behavior and its connection with the concept of rationality.

Important advances in our understanding and in developing effective tools might derive from the interaction among different disciplines such as economics, psychology and computer science.

In particular while the interaction between economics and psychology has generated the by now well established field of behavioral economics, we think that there many new potential gains from the confrontation between economics and computer science. A joint field of economics and computer science is emerging from two converging intellectual needs: on one hand computer science is becoming increasingly important for economists working with big data to address complex questions, on the other hand computer scientists increasingly are expected to understand the economics of computing systems, as well as the informational and strategic interactions between people, businesses, governments, and electronic agents.


Methods:

Our approach is to follow both a theoretical and an empirical method to construct a general framework that allows mutual understanding among different disciplinary approaches. The starting point should be the development of a common language and for this reason, we ask all the participants to try hardly to interact using a language as plain as possible. Of course, this does not mean to relax scientific requirements, but to struggle to reach the widest possible audience.

 

The project will be organized along two complementary practical strategies, a series of informal periodical seminars and some workshops.

The informal periodical seminars will be as informal as possible, so that speakers might present readings that might stimulate interests, open projects, open problems, work in progress and even working papers. However, all the presentations should be tailored to the multidisciplinary audience. Of course, some presentation might be more technical, and will be explained in the call for interests. The seminars are scheduled every two weeks on Thursday from 1.30 to 3.30 p.m. and they will be itinerant between Bicocca and Politecnico, and possibly in the future they might involve other Milan universities as well, called Permanent Itinerant Game Theory Seminars, or PIGS.

The workshops will involve up to four papers to be presented in half day, and they will be scheduled every four months, more or less, and will be more focused on specific topics and, generally, on more structured works with respect to PIGS.

In general, we believe it is particularly important to try to drag into these informal seminars as many persons as possible, coming from the most different disciplines as possible.


Moreover, we will publish our research on international scientific journals.

Approach:

Game theory, psychology, computer science


Initiatives: