Using Hidden Markov Models to Uncover Underlying States in Neuroimaging Data for a Design Ideation Task

Abstract

AbstractRecently, design researchers have begun to use neuroimaging methods (e.g., functional magnetic resonance imaging, fMRI) to understand a variety of cognitive processes relevant to design. However, common neuroimaging analysis techniques require significant assumptions relating temporal and spatial information during model formulation. In this work, we apply hidden Markov Models (HMM) in order to uncover patterns of brain activation in a design-relevant fMRI dataset. The underlying fMRI data comes from a prior research study in which participants generated solutions for twelve open-ended design problems from the literature. HMMs are generative models that are able to automatically infer the internal state characteristics of a process by observing state emissions. In this work, we demonstrate that distinct states can be extracted from the design ideation fMRI dataset, and that designers are likely to transition between a few key states. Additionally, the likelihood of occupancy within these states is different for high and low performing designers. This work opens up the door for future research to investigate the patterns of neural activation within the discovered states.