PhD Project 16
The Game of Language: Complex Communication and Mental States
Humans can communicate about factual states of the world (`there is no apple in the basket’), but can also convey and interpret information about mental states (`she does not know there is an apple in the basket’). Everyday communication largely depends on this ability to represent, to strategize, and to act upon information about mental states of other people. Using interactive game-theoretical scenarios, this project combines two levels of investigation: logic-based models to capture the structure and the complexity of reasoning about mental states underlying communication, and neurocognitive research to define the neural mechanisms supporting those complexity demands. In particular, this project will explore different ways of modeling pragmatic and non-verbal communication, to study the structure of the cognitive capacities that enable people to engage in such forms of communication. It will investigate to what extent higher levels of reasoning (about mental states) are helpful or necessary to engage in these forms of communication.
This project brings tools from the formal study of logic and computer science to cognitive neuroscience of human communication. The fact that it investigates how much higher-order knowledge of what the other participants of the conversation believe and desire is required for successful communication is highly innovative.
A Dynamic Epistemic Logic model was created to model referential language games, which are communication games in which communicators are hypothesized to coordinate their (communicative) behaviour by means of Gricean reasoning. Our model explicitly models the qualitative reasoning that could lead to such coordination, and it models how many levels of reasoning are needed for a given instance to reach successful communication. The predictions of the model were compared to those of existing models. Currently a behavioural experiment is being set up, to test the predictions of the model with experimental data. In addition, a set-up was made to model the Tacit Communication Game using the Iterated Best Response model, in which the TCG is modelled as a signalling game. This model will be used to run simulations in 2017, and the outcome of the simulations will be compared to existing data of human performance.