Data Science Research Group

Cognitive & agent modelling

Members of the CALPS group carry out research on cognitive and agent models of human performance in higher-level cognitive domains such as attention, memory, problem solving and decision making.  In the field of cognitive modelling they tend to use cognitive architectures such as ACT-R to more broadly address the rich and context dependent strategic control of cognitive and perceptual-motor processes in tasks that are often complex, dynamic and interactive.  Members of the group aspire to develop empirically grounded models that reproduce and predict human data.

Examples of research

Recent cognitive modelling research investigated information processing factors contributing to individual difference on the Attentional Network Test (Fehmida Hussain). Models were developed of adults, children of different age ranges as well as patients with Alzhiemers disease and mild Traumatic Brain Injury. The models implemented in the ACT-R architecture were able to simulate the three different functions of attention networks: alerting, orienting and executive control. This allowed investigation of which functions of attention could better account for individual differences in performance on the task. For example, a recent model of patients with traumatic brain injury suggested that their performance resulted from an impairment to the orientation network.  

A current project has been investigating the adaptive selection of strategies exibited by participants solving fire-fighting problems in a complex and dynamic microworld environment (Alberto De Obeso Orendain).  The model implemented in ACT-R has been able to capture properties of context dependent strategy selection and the adaptive tuning of strategies observed via human protocols. The research is also adressing how factors determining the processing of rewards may account for individuals differences in performance. 

Current research is also investigating the utility of agent models in predicting migration in response to climate change (Christopher Smith). The system models individuals using virtual agents in a simulated geographical environment which can undergo climate changes. Agents can perceive climate change, interacts with other agents and make decisions to migrate to new geographical locations. Decision rules and classes of parameters of agents are based on a conceptual cognitive model of internal and external factors that contributes to decision making in individuals. The model allows specific parameter values of agents to be implemented based on real empirical data which is currently being explored using data collected from a community in Burkina Faso.