ABSTRACT

This project aims at studying, developing, and testing computational models devoted to detect, recognize and interpret activities, events, and abnormal behaviours in a scene or a situation. In particular, the main objective is to discover hidden purposes underlying particular situations, that is, purposes which are not clearly evident from the observations, or are deliberately masked or immersed in other information to be kept secreted.

This will be accomplished by designing novel computer science methods integrated with theories and findings of social sciences.

The general objective is the analysis of social interactions, which is an issue investigated by the social sciences, psychology and sociology in particular, to understand all aspects of behaviour, both individually and in groups. Unfortunately, computational systems able to detect or predict the human behaviour have been barely exploited by taking into account social science findings. Therefore, we expect completely novel insights.

The core of the project is the design of computational methods able to detect, recognize, predict, in short, to model behaviours and situations. To this end, the project will also aim at constructing behavioural ontologies by considering a synergic application of probabilistic graphical models, neural-based approaches, statistical methodologies, artificial intelligence techniques, and hybrid methods as well. Moreover, social psychology, experimental psychology, forensic psychology, micro-sociology and related disciplines are definitely the inner behavioural counterpart to be exploited to design, test, and validate the novel computational methods, and will be used to achieve the appropriate ontological representations that would enable to model and predict behaviours.

The project will focus on several scales of reasoning and associated complexity which are related to the size of the monitored environment, operative conditions, and sensors/data to be used. Several situations will be considered, either engaging a huge mass of people in a large area, either more manageable situations involving a small group of persons in a confined (although still large) area, like meeting rooms, offices, or teaching rooms, and, finally, more confined situations including few people engaged in tactical/strategic board or card games.

Therefore, the proposed approach is holistic and multidisciplinary in that it will try to find the roots of (hidden) behaviours at different scales, attacking the problem using several methodologies, and taking inspiration from theories and findings of social psychology, sociology, and, in broad sense, all behavioural sciences.

More in general, this project entails a fundamental aspect of computer science, which has been little addressed so far, that is, the possibility to design a novel branch of computational science methods able to account for social science findings. This may have great implications in the progress of Information Technology (IT) as may support the development of future IT systems closer to human needs, feelings, and emotions, or, in other words, the design of human behaviour-aware devices.