Cognitive Load Dynamics in Generative AI-Assistance: A NeuroIS Study

This research investigates the real-time effects of generative AI tools, such as ChatGPT, on cognitive load during knowledge work tasks. Using NeuroIS methods (specifically, wearable EEG), the study aims to develop an understanding of how AI-assistant interactions affect cognitive demands. To this end, we extend cognitive load theory by modelling a relationship between the timing and type of AI assistant invocation and cognitive load, taking into account moderators such as task complexity and importance. For the corresponding experiment, first results from a pilot study show that AI assistants do not generally reduce cognitive load compared to traditional internet searches during a ML modelling task, indicating that they can also increase load. Based on these conceptions and findings, we derive important next steps for larger, subsequent data collections. Ultimately, this research aims to optimize AI assistance designs, by enriching them with a general understanding of cognitive load dynamics during complex work.