Physiological correlates for an agent-based computational model

Using a biofeedback system and agent-based paradigm, we developed a computational model to simulate social and economic phenomena considering emotional rules. In agent-based, we create a computer program containing program parts representing artificial agents, and we shape these agents in an environment and endow them with some rules. Then, we let them interact with each other over time in the socalled agent-based simulation, building in this way an artificial laboratory in which we can investigate many phenomena. Our contribution was to bring human beings in agent-based simulations through physiological responses of subjects at relaxed and stressed conditions. We attached the following sensors to 30 subjects: Two EEG (Electroencephalography), positioned in correspondence with the orbitofrontal cortex, one Blood Volume Pulse (BVP), one Galvanic Skin Response (GSR), and one thoracic respiration sensor. These sensors were applied to subjects who submitted to audiovisual stimuli that were designed first to relax them, then to engage them, and finally to stress them. Frequently, other authors have considered human beings in simulations through the use of avatars (i.e., agents that are fully controlled by human beings) interacting with artificial agents. Using avatars in a simulation, however, requires the inclusion of many variables all at once in the model. In fact, an individual behaves and makes decisions and choices based on many factors and variables that, at the moment, cannot be included in an agent-based model; an individual will consider many strategies that we cannot totally understand, and, even if we can understand something, this cannot be divided into the thousands of variables required. Therefore, we decided to try another way to take these variables into consideration. At the present time, we have many instruments that can help us obtain some of the variables that represent a few elements of a human being’s decision-making, behavior, and choices. For example, we could use biofeedback, as explained before, to obtain statistical data. In this way, we obtain signals from human beings under stress or during relaxation, and we can consider the status of a subject using a few physiological variables that can then be inserted into an agent-based model.