Riccardo Gallotti

The decision process can be modelled from several different perspective. They range from the biologically inspired recurrent spiking neural networks to the decision field theory, developed in economics to describe decisions under uncertainty. When the choice is between only two alternatives, decision times and error rates for all models can be re-conducted to the first passages at a barrier of a 1D diffusion process with a drift. 

This Drift Diffusion Model has been widely used for the description of perceptual tasks. Here we lift it beyond its original scope and use it to explain data from a large Game Theory experiment where human subjects made in total more than 27.000 decisions between cooperation and defection in a multiplayer Prisoners Dilemma. These strategical decision take in average a significantly larger time than perceptual choices and are associated to a different area of the brain. Nevertheless, we find that the Drift Diffusion Models fits extremely well the distribution of decision times. This shows that the model is much more universal than was originally thought and suggests that its application can be generalised to any binary choice.

For the first time, we have a tool to describe precisely the learning process in game theory experiments by tracking the evolution of the two parameters of the Drift Diffusion. Through this lens, we can observe the emergence in our experiment of cognitive biases. Indeed, people progressively develop a confirmation bias towards repeating the same action. Moreover, although initially people’s intuitive decision is to naively cooperate, over time they start to have less doubts about defection than about cooperation.

Biografía

Riccardo Gallotti is an italian physicist from the University of Bologna. He currently works at the CEA of Saclay France with M. Barthelemy on this issues: Mobility, Travel Behavior, Transportation, Complex Systems, Data Science.