Representing and Modeling Group Option-Generation Process

Generating options is crucial to making good decisions. Prior research has designed experiments to investigate how different interventions (e.g., value-focused brainstorming) affect the quantity and quality of the generated options. We propose a novel empirical method that characterizes the group option-generation process in two steps: first is to use natural language processing to represent the cognitive space of a group based on their conversation transcripts; second is to assess the discussion dynamics, e.g., inclinations of exploration versus exploitation, with a multi-dimensional Hawkes process. By applying the representation and modeling method to the brainstorming stage of a high-school product design contest, we identify three reference types of group decision-makers – mechanic, propeller, and thinker, and estimate each team participating in the context as a mixture of the three types. We further conduct model-based analysis on how the mixing configuration affects team performance in terms of their navigation strategies in the cognitive space. Finally, we report a case study on applying the proposed method to test a particular intervention, i.e., asking subjects to think about objectives beforehand, in a brainstorming exercise discussing solutions to improve student life satisfaction at our university