Cognitive network structure of Human Virtual Interactions: recent experimental results (March 2012)

In recent times there is a growing interest in measuring the behavior of human communities with the goal of forecasting its collective behavior on a statistical ground, or even be capable
of simulating the collective dynamics starting on individual basis. However, many psychological/evolutionary studies reveal that the human structures have certain characteristic
dimensions. In particular, structures that play an important role is the local community, of the order of Dumbarʼs number, and the small group, defined as the “natural” dimension of a
group of interacting people, of the order of 10-12 individuals. As Dumbar studies suggest, the main social problem to be solved by humans is that of understanding the characteristics of
the individual composing its social community, and the relationships among them. This task is endless: people change in time, new members may be introduced into the community,
others may leave. The mechanism by which this task is carried out is through almost useless chats and interactions, occurring in small groups. These groups are rather dynamical
entities, and behave differently according with the additional task proposed. A free chat in a cafè is for instance different by from a workgroup meeting, even though there are indeed
many common characteristics. It is therefore extremely interesting to try to explore the strategies used by people to interact with others, characterize and classify them. It is a common
experience that the basic relationships among people in a group can be inferred even without analyzing the messages exchanged, as can be noticed when observing a group of
foreigners chatting. On the other hand, human communication is deeply based on subtle non-verbal messages, that are difficult to be captured even with high-speed cameras, and are
anyhow quite difficult to be extracted from records. We therefore settled a virtual chat room, using a browser interface, furnishing a small amount of nonverbal tools like emoticons, mood,
and a virtual room where actors may move themselves. We analyzed the network structure of communications and the “social representations” or “affinity” that anyone has about the
others (using a “social room” where each participant is asked to place otherʼs symbols), when participant are faced with two task: a “blank modality”, where people were free to chat, and
a “task modality” where they were asked to perform a collective work. A concept that may help in understanding the group dynamics is that of pshychological field. It is a tentative
application of physics-like laws of motion to psychology (although only at the level of analogy) performed by Lewin in the ʻ50s. The idea is that present and past interactions determine
“forces” that are felt by an individual as it they were a field: by changing his internal status one can move according with these forces. The psychological field may encompass two widely
used concepts: the one of “social distance” and the one of “affinity”. The first represent the distance between two individual states, while the second is a record of past interactions.
We hope to be able to capture some of this information through the use of the “social room” tool that we included into our instrument. There are many suggestions that the task proposed
determines the network of communications, we would like to explore if the reverse is true, i.e., if one can infer the type of task by analyzing the structure of the communication network.
We have analyzed standard quantities like Derrida coefficient, triangles, clustering coefficient and activity indicators. We have found that the clustering coefficient distribution and the
Derrida spectrum can be good indicators of the collective communication state of the network. The distribution of these structures at the end of the session reveals the collective
structure of the network; the differences of these indicators between the tasks are evident. The variance analysis of the variables related to the communication sphere also shows
important differences between the two modalities. The t-Student analysis shows that in the blank modality (unconstrained) the affinity assessment seems to be unrelated with the
semantic contents of the messages, since their dynamical behavior is strongly correlated. Conversely, in the topic modality, the predictive value of the quantities related to communication
is low. This behaviour suggests the existence of two different set of cognitive strategies (or heuristics) adopted by the subjects in order to face the task, and responsible for the
emergence of the cognitive network structures. In addiction a structural analysis based on the triangle density appears as able to capture critical features of a dynamical cognitive
network such as: the metastability of an equilibrium state for the system and the effect of the cognitive constraints determined by the task on the network structure.
Finally new results coming from early experiments finished at the beginning of March 2012, are going to be presented and discussed. The previous two modalities (i.e. Blank and Topic)
will be compared with the third one, named as "Voting" modality, which introduces a frustrated game into the task.