Projects



Cognitive Modeling of Epidemics

Brief Presentation


RECOGNITION Tri Partite Model

Brief Presentation


We have developed a chatline environment that allows the recording of the interactions among a group of ten participants. We have tried to arrange an environment that mimics that of an unstructured group of stranger participants. Each participant has at his disposal a console with two textual windows, one for communicating with the rest of participants in a public way, and one to communicate to a selected subset in a private way. This should model loud conversations and whispering. Since people do not see each other, we have included two ”radars” in the interface, in
which the symbols representing other participants may be placed (1). One radar is public, seen by others, and user can only move his own symbol. The other is private, and one can move all others symbols while his own is always at the center. In this way we are trying to offer an equivalent to external non-verbal communications (the public radars) similar to changing place in order to be closer to a given person, and a mnemonic aid (the private radar) for the representation of others’ identities and their perceived social proximity, as seen by each user. In order to corroborate this interpretation, messages coming from a given individuals are darker is the individual is close in the public radar, and vanishingly clear if she is far. Finally, each message can have a ”mood”, represented by a a small icon with thumb up, down or neutral (2). This should condense the non-verbal content of a message (as usual in textual chats).




Beyond their common use for interpersonal communication, chatlines (also chat-rooms) can be formalized as dynamic systems with heuristics. We have studied chatlines in the framework of social networks. The design and data analysis of chatlines opens a new interesting research direction in social network studies. It provides the opportunity of studying the dynamics of human social behaviour in experimental ’controlled’ (or nearly controlled) conditions. Our study aims to point out both the analogy with physical systems of interacting objects and the social network emerging properties linked to the existence of different communication patterns and usage of different heuristics in the participants. We describe guidelines for effective implementation of a chatline in controlled experimental conditions. We identified several parameters which represent meaningful statistical estimators of the activity of the network and we computed the correlation of these parameters and measures of network statistics.[Pubblications:
1]


Detecting communities is a task of great importance in many disciplines, namely sociology, biology and computer science, where systems are often represented as graphs. Primate communities are strongly social, and therefore we have developed sophisticated tools for detecting the various degree of relationship. These tools are quite slow, since in primate a large part of time is devoted to exploration and communication, and require large computational efforts and large brainsCommunity detection is linked to clustering of data: many clustering methods establish links among representative points that are nearer than a given threshold, and then proceed in identifying communities on the resulting graphs. In particular, we deal with the task of identifying communities in an existing graphs, using a local algorithm and not relying on global quantities like betweenness, centrality, etc. An individual is simply modeled as a memory and a set of connections to other individuals. We explore two different approaches: in the first, information about neighboring nodes if propagated and elaborated locally, but the connections do not change. In the second approach, information is not elaborated while it is the wiring that is varied with the result of directly connecting to a “central node”. Both processes can be considered implementations of the availability heiristic, which is simply the assumption that the most vivid or easily recallable information give an accurate estimate of the frequency of the related event in the population.