Can a Model of Mental Math Problem Solving Be Applied to Cyber Capture the Flag Problems?

Abstract

Problem-solving processes typically involve several serial stages or steps. At a fine grain size, the nature of the steps required may be domain- (or even problem-) specific. At a coarser level, we could assume a that common sequence of general stages applies across the problems of interest. When someone solves a problem in his/her head, how-ever, it is difficult to externally discern the nature of these problem-solving stages, and the timing of inter-stage transitions. An MVPA-HSMM modelling approach fed by neuroimaging data revealed that a 4-stage model was appropriate for a set of math problems, and provided estimates of the time invested in each stage on each trial. We could thus determine which stage durations were affected by different experimental and individual factors (including level of expertise). Such model-facilitated insights would also be valued in the context of Cyber Capture the Flag problems, which are used to train and assess cybersecurity skills. Can a stage model usefully apply to differentiate among different strategies, and inform assessments of expertise in this context? If so, for individuals solving problems at a computer rather than inside a scanner, is there a sufficiently rich source of data (e.g., the state of the computer rather than the state of the solver’s brain) to support a bottom-up/data driven modelling approach to determine the boundaries of these stages ‘automatically’ or will the boundaries need to be determined more manually.