My first research question is whether people are able to judge the reliability of their partners in the absence of objective feedback and what are the likely underlying mechanisms. For this purpose, I am using a modified advice-revise paradigm where people make a decision and then are offered the possibility to revise their own judgement after knowing the opinion of a virtual partner. It is known that people are affected by an ego-centric bias- meaning that they overestimate their opinions and underestimate the opinions of others. After repeated interactions with the same partners however, people change the weight given to others’ decisions allowing them to be more or less influential on their own choices. By comparing human performance with an optimal Bayesian observer, we can identify differences and similarities and possibly understand the neural mechanisms underlying these processes.
My second research question focuses on how information is transformed during ecological social interactions and how we can fully describe information manipulation during a interaction. For this purpose, we are bringing together two (or more) people to the lab and make them interact through a computer interface. Later, their behaviour is analysed both at the individual and collective level in order to understand how interaction affects what the single individuals do and think and whether a machine can simulate the ecological behaviour of a social partner. Although interaction among multiple agents gets exponentially complex, understanding it through geometrical representations like the Opinion Space allows us to understand it and predict it.
My third line of research started one Friday afternoon to have some academic fun. Inspired by the Human Dynamic Clamp paradigm that pairs one human agent with one virtual one I developed a virtual partner that dynamically changes its confidence based on the human current expressed opinion. This allows for a real-time social interaction between a human and a machine, but has the advantage of having total control over the parameters governing the machine behaviour. These parameters can thus be modified and experimented upon in order to create a good replica of human social dynamics.