Exploring latent states of problem?solving competence using hidden Markov model on process data

The response process of problem-solving items contains rich information about respondents' behaviours and cognitive in the digital tasks, while extraction is a big challenge. aim study to use data-driven approach explore latent states state transitions underlying reflect test-takers' behavioural patterns, investigate how these could be associated with performance. We employed Hidden Markov Modelling identify test takers' hidden during compared frequency and/or between different performance groups. conducted comparable studies two focus on US sample that was collected PIAAC 2012, examined correlation those frequencies from items. Latent them were identified found significantly by groups correct responses both more engaged tasks often efficient tools solve problems, group incorrect likely shorter action sequences exhibit hesitative behaviours. Consistent patterns across This demonstrates value based HMM better understand transmissions underneath observable complex tasks.